Analysis report

NBA Player Salary

Data from R package wooldridge (CRAN), published under https://diafunc.com/licences/view-only.

Target
Wage
Analyzed
2026-11-07 13:15 UTC

Summary

The model predicts Wage and explains 56% of the variation, relying most strongly on Avgmin. In this data, moving Avgmin from 19.1 to 33.4 raises the model's output by about 772 on average, with everything else as observed. However, a straight line through the data explains 67% of the variation, so the model comes in 11 points lower on variance explained, and the margin is small.

The model did not beat the baseline, so its formula adds no value beyond a simple regression. Complexity is below the midpoint, so the formula is not difficult to read from a technical perspective. Given these results, reviewing the modeling approach or inputs is warranted before using this model for decisions. The next step is to review the modeling process and validate alternatives.

Formula

Wage = 0.07 × Agesq × Avgmin + 150Wage=0.07×Agesq×Avgmin+150
Wage = 0.07 × Agesq × Avgmin + 150
2 terms · additively decomposable

Held-out diagnostics

TargetQualityErrorR²MAESearch rowsGeneralizationRows
Wage50.98%43.97%0.5603479.2195generalizing54
  • Wage · Quality 95% CI 30.86% to 64.53%
  • Wage · Error 95% CI 31.82% to 62.01%
  • Wage · R² 95% CI 0.3799 to 0.6818
  • Wage · MAE 95% CI 384.5 to 580.9

Concerns

CRITICAL model

The model loses to a straight line through the data on `Wage`

The model explains 56% of the variation on the held-out rows, where a straight line through the data explains 67%: 11 points less. The reference predictor is the better choice unless a readable formula is the point. A plain regression on the same inputs does as well or better, so the inputs do carry signal; the search did not find a better formula in the time it had. Row by row the two are about level.

WARNING data

`Wage` differs where `Draft` is missing

In the 29 rows where Draft is missing, Wage averages 523; in the 240 rows where it is present, 1,533. The difference is -1,009 (95% interval -1,198 to -821), 1.06 standard deviations of Wage. The gaps are not spread evenly over the values of the target. A model is computed on rows with a value in every input, so it sees a target that differs from the table's, and where a value is missing is information it does not get. A column that marks the rows where the value is missing keeps it.

INFO model

Residuals grow with the predicted value, more slowly than it

Across 54 held-out rows split into thirds by predicted value, the residual standard deviation in the top third is 807, against 442 in the bottom third, 1.82 times as large. The bottom third holds predictions up to 944 and the top third predictions from 1,657. Relative to the predictions, the typical miss is 62% of the value in the bottom third and 38% in the top third: the errors grow with the value, but more slowly than it, so relative to the value the model is more accurate in the top third than in the bottom third.

INFO model

`Lwage` reproduces `Wage`

Spearman's rho = 1 (Pearson r = 0.894) between Lwage and the target Wage. A match this exact is what a copy or a rescaled copy of the target looks like. The coefficients read the same for a strong input and for a leak. If Lwage is known before Wage exists, it is the strongest input and the quality stated in this report is real. If it was computed from Wage or recorded after it, a formula that reads it will not hold on new data. The model does not read this column, so its predictions do not depend on the answer.

Analysis

The report opens with the Headline and the Concerns, followed by the Discovered Relations when the analysis has targets and by Questions & Answers when questions were asked. Three sections carry the detail: Model presents one model per target, Implications states where the models hold and what to weigh before relying on them, and Data describes the table as supplied, column by column, for reference.

Generated 2026-11-07 13:15 UTC, analysis 01760490-48b8-11f1-8087-d378640fe787, 1 target, time budget 1 h 0 min, compute budget 500 credits, 269 rows. The Provenance & Audit chapter holds the full record.

Validity

Validity: valid. Nothing in the data or the formula speaks against relying on this result.

How to read this

Whether this result is one to rely on. Quality says how closely the formula follows the data; validity says whether that closeness means anything. A formula that only restates its target, or that reads a column which already holds the answer, can reach a high quality and still say nothing new about the data.

What it found

Wage starts from a base of 150 and rises with Avgmin and Agesq.

  1. Avgmin carries the model of Wage. Avgmin accounts for 100% of what the formula explains.
  2. The model's output rises with Avgmin. With the other inputs at typical values, the formula gives 297 at Avgmin 2.89 and 2350 at 43.1. The change is not even across the range.
  3. Points is not in the formula. It has the third-strongest correlation with Wage in the table (0.66). Rows with more of it have more Avgmin (correlation 0.89), and Avgmin is in the formula.
  4. The formula does worse than a straight line. The formula explains 56% of Wage; a straight line explains more, 67%. That was measured on 54 rows it never saw. The formula typically misses Wage by 479.

Summary

What to do next: The model loses to a straight line through the data on Wage.

6 chapters, each headed by what it found.

S.1 Avgmin carries the whole formula

What the formula relies on. Wage rises as Avgmin rises.

Measured on 19 of the 195 rows the model was found on, which the search set apart for it. On so few rows the order of the inputs is indicative, not settled. The model's output rises with Avgmin on every one of the 176 measured rows (exact derivative). The model's output rises with Agesq on every one of the 2,992 steps between observed values. Moving Avgmin from 19.1 to 33.4, its first and third quartile, with every other input as observed raises the model's output by 772 on average (by between 441 and 1,682 on single rows). These are measurements on the rows the search trained on, inside the observed ranges; outside them nothing is claimed.

Table 5.1.3 · Which inputs matter most
InputShare of importancePermutation importanceSensitivityResidual correlationOccurrencesModel's output as the input risesChange, first to third quartile
Avgmin100%1.1784753.23-0.1241Rises (176 of 176 rows)+772.3 (19.1 to 33.39)
Agesq-0.094901.7890.29261Rises (2992 of 2992 steps)+486.6 (625 to 900)
Rows the search trained on, measured
176
N rows
19
Inputs the formula does not read
Age, Allstar, Assists, Black, Center, Children, Years played in college, Draft, Exper, Expersq, Forward, Games, Guard, Marr, Marrblck, Minutes, Points, Rebounds

The formula decomposes into 2 additive terms: clean separable structure. Each curve below shows one term's response to its free inputs, with the other inputs held at the training median. Steep curves carry the prediction; flat curves are decorative.

Additive
Yes
Rows the search trained on, measured
176
Terms in the formula
2
Notation
0.07 × Agesq × Avgmin + 150
Target
Wage

S.2 A straight line does better here: R² 0.56 against 0.67

How well it holds. The formula explains 56% of Wage; a straight line explains more, 67%.

How this was checked

  • Measured on 54 rows the formula never saw (22% of 249); the other 195 built it.
  • Compared on the same rows against always guessing the average, always guessing the median and a straight line.
  • Intervals from 1,000 bootstrap resamples of those rows.

R² of 0.56 on 54 rows the search never saw (95% interval 0.38 to 0.682): weak or better. The model captures only a small part of the variance; either the relevant inputs are missing, or the target depends on interactions the formula did not recover. Mean absolute error is 479 on the target's scale.

Error
43.97% (95% CI 31.82% to 62.01%)
MAE
479.2 (95% CI 384.5 to 580.9)
N
54
Training rows
195
Quality
50.98% (95% CI 30.86% to 64.53%)
R²
0.5603 (95% CI 0.3799 to 0.6818)
Rows found on
195
Target
Wage

The model does not beat the simple yardsticks: a straight line through the data is ahead by 11 points of variance explained on the 54 held-out rows. The model's complexity is not buying anything yet. A plain regression on the same inputs does as well or better, so the inputs do carry signal; the search did not find a better formula in the time it had. Of the 54 held-out rows where the two differ, the model is closer on 35% (95% interval 24% to 49%).

Best yardstick's score
0.6708 (95% CI 0.498 to 0.7824)
Best baseline R²
0.6708
Effect size d
-0.2908 (95% CI -0.5432 to -0.0391)
Effect size d magnitude
small
Margin over the best yardstick
-0.1105 (95% CI -0.2245 to -0.007842)
Margin's interval stays above zero
No
Model MAE
479.2 (95% CI 384.5 to 580.9)
Model
0.5603 (95% CI 0.3799 to 0.6818)
Model R²
0.5603 (95% CI 0.3799 to 0.6818)
Held-out rows
54
Held-out rows without an answer
0
Target
Wage
Rows where the model is closer
35.19% (95% CI 23.82% to 48.52%)
Rows compared
54
Rows level
0
Regression yardstick trained with gaps filled (too few complete rows)
No
Held-out rows with a gap filled for the regression yardstick
9
Training rows of the regression yardstick
195

The model explains held-out rows about as well as training rows (held-out R² 0.56 on 54 rows, gap between training and held-out R² -0.127): it generalizes. The structure the formula encodes carries from the training rows to the held-out rows.

Generalization gap
-0.1271
N test
54
Target
Wage
Test error
479.2 (95% CI 384.5 to 580.9)
Test R²
0.5603 (95% CI 0.3799 to 0.6818)
Train error
0.5964
Train R²
0.4332
Verdict
Generalizing

S.3 No range of its inputs where it is markedly weaker

Where it is weak. The largest miss on a held-out row is 1470 (recorded 3125, predicted 1660).

Segment children = no performs in line with the aggregate (37 held-out rows): segment R² 0.454 against aggregate R² 0.56 (gap -0.106). No quality drift detected.

Table 6.2.1 · Wage by Input Range
SegmentRowsR²Gap to all rowsVerdict
Children = false370.4539-0.1065In line with all rows
Center = false510.5373-0.0230In line with all rows
Black = true480.5371-0.0233In line with all rows
Allstar = false470.5279-0.0324In line with all rows
Guard = false310.6273+0.0669In line with all rows
Marr = false330.5089-0.0514In line with all rows
Marrblck = false360.5582-0.0022In line with all rows
Years played in college = q4430.5507-0.0097In line with all rows
Assists = q4130.2306-0.3297Too few rows to judge
Lwage = q411-1.2218-1.7821Too few rows to judge

The first 10 rows of 24. The other 14 are in the notebook.

Aggregate MAE
479.2
Aggregate R²
0.5603
Measure
r2
Target
Wage

The worst held-out row was missed by 1,468 (target's scale). The 10 rows with the largest absolute residuals out of 54 held-out rows are shown. Each card breaks the failure down into per-term contributions and input values, so the failure mode is inspectable. A large residual with one concentrated term contribution means one term got this row wrong; a large residual with balanced contributions means a structural mismatch.

Table 5.1.13 · Worst Predictions
ActualPredictedResidualAgesqAvgmin
31251657+146862534.45
35872192+1395108926.79
40002712+128896138.08
29251692+123357638.23
2000867+113310899.4
11202202-108284134.86
1501225-107567622.71
11002152-1052108926.26
1501063-91367619.3
1498626+87257611.81
Additive
Yes
Expected coverage
0.8673
N extrapolated
0
Held-out rows
54
Held-out rows for the interval
54
Rows inside the interval
51
Residual P5
-961.8
Residual P95
1252.7
Residual std
612.5

0 of 54 held-out rows (0%) lie further from the training rows than 99 in 100 rows like them would, measured on the 2 inputs the model reads (agesq, avgmin). That is a small minority; the held-out figures describe the model on rows like the ones it was trained on.

Table 6.3 · Rows unlike the ones the model learned from
TargetInputs measuredLineHeld-out rowsBeyond the lineShare
WageAgesq, Avgmin3.0355400%
Table 6.3 · Score bins · 24 of 30 rows shown
CountFromTo
000.121
10.1210.243
20.2430.364
10.3640.486
00.4860.607
40.6070.728
20.7280.85
40.850.971
40.9711.093
61.0931.214

The first 10 rows of 24. The other 14 are in the notebook.

S.4 Not ready to use yet: 1 critical concern

What to watch. First: The model loses to a straight line through the data on Wage.

Table 3 · Concerns
SeveritySectionConcernDetail
CriticalModelThe model loses to a straight line through the data on WageThe model explains 56% of the variation on the held-out rows, where a straight line through the data explains 67%: 11 points less. The reference predictor is the better choice unless a readable formula is the point. A plain regression on the same inputs does as well or better, so the inputs do carry signal; the search did not find a better formula in the time it had. Row by row the two are about level.
WarningDataWage differs where Draft is missingIn the 29 rows where Draft is missing, Wage averages 523; in the 240 rows where it is present, 1,533. The difference is -1,009 (95% interval -1,198 to -821), 1.06 standard deviations of Wage. The gaps are not spread evenly over the values of the target. A model is computed on rows with a value in every input, so it sees a target that differs from the table's, and where a value is missing is information it does not get. A column that marks the rows where the value is missing keeps it.
InfoModelResiduals grow with the predicted value, more slowly than itAcross 54 held-out rows split into thirds by predicted value, the residual standard deviation in the top third is 807, against 442 in the bottom third, 1.82 times as large. The bottom third holds predictions up to 944 and the top third predictions from 1,657. Relative to the predictions, the typical miss is 62% of the value in the bottom third and 38% in the top third: the errors grow with the value, but more slowly than it, so relative to the value the model is more accurate in the top third than in the bottom third.
InfoModelLwage reproduces WageSpearman's rho = 1 (Pearson r = 0.894) between Lwage and the target Wage. A match this exact is what a copy or a rescaled copy of the target looks like. The coefficients read the same for a strong input and for a leak. If Lwage is known before Wage exists, it is the strongest input and the quality stated in this report is real. If it was computed from Wage or recorded after it, a formula that reads it will not hold on new data. The model does not read this column, so its predictions do not depend on the answer.

The analysis ran with the configuration below. The checks and the verdict of this chapter are read against it.

  • Targets: 1 target: wage
  • Rows: 269 rows
  • Per-Segment Predictions: not produced, because no segment column was chosen
  • Fairness: compared across 8 categorical columns (allstar, black, center, children, forward, guard, marr, marrblck); which of them are protected attributes is not decided here
  • Out-of-Distribution: held-out rows are scored against the training rows; predictions made outside this report are not scored
  • Provenance & Audit: recorded in the chapter of that name

Across 6 checks (3 pass, 2 warn, 1 fails, 1 critical concern), a check fails and a critical concern is open: Not ready to use yet. The failing checks below and the critical concerns in the Concerns chapter need addressing before the predictions are used.

Table 6.6 · Deployment Readiness
NameStatusDetail
GeneralizationPassquality on the held-out rows matches quality on the rows the model was found on
Prediction Interval CoverageWarntoo few held-out rows to judge coverage, largest gap 13.0 percentage points
Baseline Lift for WageFailloses to the best baseline, R² lift -0.11
Model ConstraintsPassevery held-out prediction is finite and within the training range
FairnessWarn8 columns compared, 4 flagged: Allstar (a group both receives smaller predictions and is systematically mis-predicted), Black (a group receives systematically smaller predictions), Center (a group receives systematically smaller predictions), Children (a group receives systematically smaller predictions)
Out-of-DistributionPass0% of held-out rows are out of distribution
N critical concerns
1
N fail
1
N pass
3
N warn
2
Overall verdict
Not ready

Column Allstar splits the held-out rows into 2 groups. The prediction-parity ratio is 0.61 (smallest over largest mean prediction), the largest absolute group residual mean is 391 (0.639 times the overall residual standard deviation). Parity gap and residual bias: the model both predicts systematically lower for some groups and errs systematically against at least one group. This is the worst-case fairness signal on this column; review before deployment. points goes closely with allstar (association 0.607 on a scale from 0 to 1): a model that reads points can tell the groups of allstar apart, so leaving allstar out of the inputs does not keep it out of the predictions.

Aggregate residual std
612.5
N columns
8
N proxies
6
Proxy threshold
0.5
Table 6.4.1 · Columns
Bias max abs residual meanColumnGroupsInterpretationN groupsN proxiesPrediction parity ratioVerdict
391.5Allstartrue, falseColumn Allstar splits the held-out rows into 2 groups. The prediction-parity ratio is 0.61 (smallest over largest mean prediction), the largest absolute group residual mean is 391 (0.639 times the overall residual standard deviation). Parity gap and residual bias: the model both predicts systematically lower for some groups and errs systematically against at least one group. This is the worst-case fairness signal on this column; review before deployment. points goes closely with allstar (association 0.607 on a scale from 0 to 1): a model that reads points can tell the groups of allstar apart, so leaving allstar out of the inputs does not keep it out of the predictions.210.6096Parity and bias
134.1Blacktrue, falseColumn Black splits the held-out rows into 2 groups. The prediction-parity ratio is 0.777 (smallest over largest mean prediction), the largest absolute group residual mean is 134 (0.219 times the overall residual standard deviation). Parity gap: at least one group receives systematically smaller predictions on average. Whether that is a property of the target itself or a model bias against the group needs checking.20.7772Prediction parity gap
13.13Centerfalse, trueColumn Center splits the held-out rows into 2 groups. The prediction-parity ratio is 0.379 (smallest over largest mean prediction), the largest absolute group residual mean is 13.1 (0.0214 times the overall residual standard deviation). Parity gap: at least one group receives systematically smaller predictions on average. Whether that is a property of the target itself or a model bias against the group needs checking.20.3789Prediction parity gap
22.53Childrenfalse, trueColumn Children splits the held-out rows into 2 groups. The prediction-parity ratio is 0.695 (smallest over largest mean prediction), the largest absolute group residual mean is 22.5 (0.0368 times the overall residual standard deviation). Parity gap: at least one group receives systematically smaller predictions on average. Whether that is a property of the target itself or a model bias against the group needs checking.20.6951Prediction parity gap
89.35Forwardtrue, falseColumn Forward splits the held-out rows into 2 groups. The prediction-parity ratio is 0.873 (smallest over largest mean prediction), the largest absolute group residual mean is 89.4 (0.146 times the overall residual standard deviation). Balanced: every group is within the 80% parity band and no group's residual mean exceeds 30% of the overall residual standard deviation. No systematic disparity was detected on this column. guard goes closely with forward (association 0.707 on a scale from 0 to 1): a model that reads guard can tell the groups of forward apart, so leaving forward out of the inputs does not keep it out of the predictions.210.8726Balanced
99.29Guardfalse, trueColumn Guard splits the held-out rows into 2 groups. The prediction-parity ratio is 0.999 (smallest over largest mean prediction), the largest absolute group residual mean is 99.3 (0.162 times the overall residual standard deviation). Balanced: every group is within the 80% parity band and no group's residual mean exceeds 30% of the overall residual standard deviation. No systematic disparity was detected on this column. forward goes closely with guard (association 0.707 on a scale from 0 to 1, and 1 more column at 0.5 or above): a model that reads forward can tell the groups of guard apart, so leaving guard out of the inputs does not keep it out of the predictions.220.9995Balanced
91.12Marrtrue, falseColumn Marr splits the held-out rows into 2 groups. The prediction-parity ratio is 0.885 (smallest over largest mean prediction), the largest absolute group residual mean is 91.1 (0.149 times the overall residual standard deviation). Balanced: every group is within the 80% parity band and no group's residual mean exceeds 30% of the overall residual standard deviation. No systematic disparity was detected on this column. marrblck goes closely with marr (association 0.802 on a scale from 0 to 1): a model that reads marrblck can tell the groups of marr apart, so leaving marr out of the inputs does not keep it out of the predictions.210.8846Balanced
122.4Marrblcktrue, falseColumn Marrblck splits the held-out rows into 2 groups. The prediction-parity ratio is 0.861 (smallest over largest mean prediction), the largest absolute group residual mean is 122 (0.2 times the overall residual standard deviation). Balanced: every group is within the 80% parity band and no group's residual mean exceeds 30% of the overall residual standard deviation. No systematic disparity was detected on this column. marr goes closely with marrblck (association 0.802 on a scale from 0 to 1): a model that reads marr can tell the groups of marrblck apart, so leaving marrblck out of the inputs does not keep it out of the predictions.210.8606Balanced

S.5 A larger held-out set would firm up the figures

What to collect next. The figures rest on 54 held-out rows.

  • Collect more rows where Draft is missing. Reason: the analysis warns that Wage differs where Draft is missing, which may signal that important variation is not captured in these cases.
  • Review and address the model’s bias and prediction parity gaps for Allstar, Black, Center and Children. Reason: fairness flags indicate parity and bias for Allstar and prediction parity gaps for Black, Center and Children.
  • Inspect and improve the model for Wage to close the gap with a straight line through the data. Reason: the model explains 56% of the variation in Wage, but loses to a straight line which explains 67%, an 11-point margin that is not settled as the margin’s interval includes zero.
  • Add or review columns that clarify how Lwage relates to Wage. Reason: the analysis finds that Lwage reproduces Wage, suggesting potential redundancy or an unmodeled transformation.

S.6 It shows how Wage moves with its inputs in these rows, not what causes it

What this model does not show. It was built on Avgmin from 2.89 to 43.1; outside those ranges it has not been checked. 58 slices of the data had too few held-out rows to judge.

Diafunc found a formula that captures the structure in the training data. That structure is correlative: it states what predicts what in this data. It does not establish that any input causes the target.

A correlative model is the right tool for predicting the target on new rows like these, for exploring which inputs carry information about the target, and for an early-stage diagnostic or screening tool.

A correlative model is not the right tool for claiming that intervening on an input will change the target, for attributing a target's values to specific causes for accountability, or for choosing between policies that act on the inputs.

The Quality by Input Range, Fairness and Concerns chapters show the most common confounding patterns in the data.

The model is correlative: it captures which columns predict the target in the training data, not whether changing one would change the other. Predictions are reliable for inputs that look like the training distribution; they are not evidence that intervening on an input will change the target. The 4 upgrade pathways below show what evidence would support a causal claim.

Table 6.5 · Upgrade pathways
NameSummary
Controlled experimentRandomize the input to be acted on, then analyze again. Randomization breaks confounding, so any change in the target is attributable to the input.
Natural experimentFind a setting where the input varied for reasons unrelated to the target, an instrumental variable. Variation of that kind supports causal inference.
Causal graphWrite out what plausibly causes what, then check whether the model's terms are consistent with the conditional independences the graph implies.
Control for confoundersIf a third variable plausibly moves both the input and the target, stratify by it or condition on it before training again. Gaps in the Quality by Input Range chapter are a clue.

Evidence

Every chapter of the analysis in full: the figures the story rests on, and everything else the analysis measured.

1 Headline

Diafunc modeled Wage from 21 columns, of which the model uses 2. The model explains 56% of the variation in Wage and relies most on Avgmin, with which the model's output rises.

  • One target was modeled; 54 rows were held out for checking.
  • A straight line through the data already explained 67%; the model fell behind by 11 points of variance explained.
  • The model loses to a straight line through the data on Wage (Critical, 1).
Best R²
0.5603
Best quality
50.98%
Best target
wage
Targets
1

2 Stakeholder Summary

The model predicts Wage and explains 56% of the variation, relying most strongly on Avgmin. In this data, moving Avgmin from 19.1 to 33.4 raises the model's output by about 772 on average, with everything else as observed. However, a straight line through the data explains 67% of the variation, so the model comes in 11 points lower on variance explained, and the margin is small.

The model did not beat the baseline, so its formula adds no value beyond a simple regression. Complexity is below the midpoint, so the formula is not difficult to read from a technical perspective. Given these results, reviewing the modeling approach or inputs is warranted before using this model for decisions. The next step is to review the modeling process and validate alternatives.

3 Concerns

Table 3 · Concerns
SeveritySectionConcernDetail
CriticalModelThe model loses to a straight line through the data on WageThe model explains 56% of the variation on the held-out rows, where a straight line through the data explains 67%: 11 points less. The reference predictor is the better choice unless a readable formula is the point. A plain regression on the same inputs does as well or better, so the inputs do carry signal; the search did not find a better formula in the time it had. Row by row the two are about level.
WarningDataWage differs where Draft is missingIn the 29 rows where Draft is missing, Wage averages 523; in the 240 rows where it is present, 1,533. The difference is -1,009 (95% interval -1,198 to -821), 1.06 standard deviations of Wage. The gaps are not spread evenly over the values of the target. A model is computed on rows with a value in every input, so it sees a target that differs from the table's, and where a value is missing is information it does not get. A column that marks the rows where the value is missing keeps it.
InfoModelResiduals grow with the predicted value, more slowly than itAcross 54 held-out rows split into thirds by predicted value, the residual standard deviation in the top third is 807, against 442 in the bottom third, 1.82 times as large. The bottom third holds predictions up to 944 and the top third predictions from 1,657. Relative to the predictions, the typical miss is 62% of the value in the bottom third and 38% in the top third: the errors grow with the value, but more slowly than it, so relative to the value the model is more accurate in the top third than in the bottom third.
InfoModelLwage reproduces WageSpearman's rho = 1 (Pearson r = 0.894) between Lwage and the target Wage. A match this exact is what a copy or a rescaled copy of the target looks like. The coefficients read the same for a strong input and for a leak. If Lwage is known before Wage exists, it is the strongest input and the quality stated in this report is real. If it was computed from Wage or recorded after it, a formula that reads it will not hold on new data. The model does not read this column, so its predictions do not depend on the answer.

How to read this

Data-quality, model-quality, and deployment-risk issues detected during the analysis appear below, ordered by severity. Critical concerns are listed first; Warning and Info entries follow.

4 Discovered Relations

Table 4 · Quality and error measured on the 54 held-out rows
TargetFormulaQualityComplexityError
WageWage = 0.07 × Agesq × Avgmin + 15050.98%6.8%43.97%

How to read this

One model per target, each a formula with its quality, complexity and error. Higher quality at lower complexity describes the data most clearly. The findings below read the models together; the per-target chapters in the Model section carry each formula, its predictions and its diagnostics.

5 Model

The model for wage relies entirely on Avgmin for its output and tracks about 56% of the share of variation explained on held-out rows. The model loses to a straight line through the data, which is flagged as a critical concern in the Concerns chapter.

Most notable in the Model section:

  • Avgmin carries most of the model (100% of its reliance on inputs)

How to read this

The Model section presents one model per target: its formula, its predictions on held-out rows, the inputs it relies on, the baselines it is measured against, and diagnostics of how it behaves. With more than one target, the Composite Model, Multi-Output and Leaderboard chapters read the models as a set.

5.1 Wage

The Model section builds one model for Wage: a formula that expresses it through the other columns. Diafunc searches the space of mathematical expressions for the formula that balances accuracy against simplicity, so the result is not a black box but an expression that can be read, evaluated and questioned. The chapters that follow show the formula, score it on rows it never saw during training, and take it apart term by term.

5.1.1 Wage Model

The model's output rises with both Avgmin and Agesq. The strongest input is Avgmin, and the formula is Wage = 0.07 × Agesq × Avgmin + 150.

Wage = 0.07 × Agesq × Avgmin + 150f:(Agesq,Avgmin)↦WageAgesq,Avgmin,Wage∈RWage=0.07×Agesq×Avgmin+150
Wage = 0.07 × Agesq × Avgmin + 150
Quality 50.98% · Error 43.97% on 54 rows the search never saw · Found on 195 rows · Complexity 6.8%
Table 5.1.1 · Constants of the formula
ValueRoleInputUnit
0.07Coefficient
150InterceptUnit of the model's output
Rows the search trained on, measured
176
Rows the search trained on
176
Held out R²
0.5603 (95% CI 0.3799 to 0.6818)
Constant base
150
Constants in the formula
2
Terms that read an input, constants multiplied in
1
Training rows
195
Held-out rows
54
Inputs the formula reads
2
Terms in the formula
2
Inputs the formula does not read
Age, Allstar, Assists, Black, Center, Children, Years played in college, Draft, Exper, Expersq, Forward, Games, Guard, Marr, Marrblck, Minutes, Points, Rebounds

How to read this

The three gauges score the model on the held-out rows: rows that were set aside before the search and that the model never saw. They are the figures to quote for how the model does on new data.

Table 5.1.1 · Wage Model
MetricMeaning
QualityHow well the model predicts on the held-out rows, marked down for the size of its formula: 100% is an exact prediction, 0% is no better than always predicting the average. It is not R², and sits at or just below it
ErrorHow far the predictions lie from the held-out values, as a share of the miss of always predicting the average: 0% is an exact prediction, 100% is no better than the average
ComplexityThe size of the formula as a share of the complexity the search was allowed: a low figure is a short formula

The figures under During the search are the same measures on the training rows. They are what the search ranked its candidates by, so they describe the search and tend to flatter the model it chose. The Generalization chapter sets the training rows beside the held-out rows, and the Baselines chapter sets the model beside its baselines. The time and credit gauges account for the computation that produced the model, not for the model itself.

5.1.2 Prediction

The model predicts Wage for any row that supplies the other columns. This chapter applies it to the held-out rows, the rows set aside before training, so that the predictions can be compared with values the model never saw. The rows themselves and the function that predicts them are kept as reference under Foundation, with the processed data.

5.1.2.1 Prediction Chart

R² of 0.56 on 54 rows the search never saw (95% interval 0.38 to 0.682): weak or better. The model captures only a small part of the variance; either the relevant inputs are missing, or the target depends on interactions the formula did not recover. Mean absolute error is 479 on the target's scale.

Error
43.97% (95% CI 31.82% to 62.01%)
MAE
479.2 (95% CI 384.5 to 580.9)
N
54
Training rows
195
Quality
50.98% (95% CI 30.86% to 64.53%)
R²
0.5603 (95% CI 0.3799 to 0.6818)
Rows found on
195
Target
Wage

How to read this

Actual Wage on the horizontal axis against the model's predicted value on the vertical axis, for the held-out rows the model never saw during training. Points on the diagonal are perfect predictions; systematic vertical offsets reveal bias, vertical spread reveals noise, and points far from the diagonal are rows the model handles poorly. With 10,000 held-out rows or more, the rows are counted instead of drawn one point each: each cell is a range of the actual and a range of the predicted value, its color the number of rows that fall in both, and the cells whose two ranges match hold the rows the model predicts well.

5.1.2.2 Residual Distribution
Bias
-7.861 (95% CI -175.2 to 152.4)
N
54
P50
-142.3 (95% CI -266.2 to 143.1)
P90
834.6 (95% CI 353.4 to 1288.4)
P95
1252.7 (95% CI 632.9 to 1420.3)
P99
1429.1 (95% CI 1067.6 to 1467.8)
Std
612.5 (95% CI 482.6 to 722.9)

How to read this

How the held-out residuals (actual Wage minus predicted) are distributed. Bias is their mean: a value away from zero means the model is systematically high or low rather than merely imprecise. The standard deviation and the percentiles describe the spread, and the tail percentiles say how large the worst misses get. Every figure carries a bootstrap 95% confidence interval; with a small held-out set those intervals are wide, and that width is itself the finding.

5.1.3 Which inputs matter most

Measured on 19 of the 195 rows the model was found on, which the search set apart for it. On so few rows the order of the inputs is indicative, not settled. The model's output rises with Avgmin on every one of the 176 measured rows (exact derivative). The model's output rises with Agesq on every one of the 2,992 steps between observed values. Moving Avgmin from 19.1 to 33.4, its first and third quartile, with every other input as observed raises the model's output by 772 on average (by between 441 and 1,682 on single rows). These are measurements on the rows the search trained on, inside the observed ranges; outside them nothing is claimed.

Table 5.1.3 · Which inputs matter most
InputShare of importancePermutation importanceSensitivityResidual correlationOccurrencesModel's output as the input risesChange, first to third quartile
Avgmin100%1.1784753.23-0.1241Rises (176 of 176 rows)+772.3 (19.1 to 33.39)
Agesq-0.094901.7890.29261Rises (2992 of 2992 steps)+486.6 (625 to 900)
Rows the search trained on, measured
176
N rows
19
Inputs the formula does not read
Age, Allstar, Assists, Black, Center, Children, Years played in college, Draft, Exper, Expersq, Forward, Games, Guard, Marr, Marrblck, Minutes, Points, Rebounds

How to read this

The chart ranks the inputs of the Wage model by permutation importance: how much the model's error rises when that input's values are shuffled across rows so that it carries no information. A longer bar means the model relies on the input more. The order says which columns the model reads most, and those are the columns where measurement quality and data collection matter most. A short bar marks an input the model barely uses, which is a fact about this model and not a verdict on the column. Importance describes the model's reliance and nothing more: an input can rank high because it stands in for something the data does not contain. Where the search measured it, an input also carries the direction in which the model's output moves as the input rises, counted on the rows the search trained on, and the change of the output when the input alone goes from its first to its third quartile; both describe the formula inside the observed range and say nothing about causes.

5.1.4 Baselines

The model does not beat the simple yardsticks: a straight line through the data is ahead by 11 points of variance explained on the 54 held-out rows. The model's complexity is not buying anything yet. A plain regression on the same inputs does as well or better, so the inputs do carry signal; the search did not find a better formula in the time it had. Of the 54 held-out rows where the two differ, the model is closer on 35% (95% interval 24% to 49%).

Best yardstick's score
0.6708 (95% CI 0.498 to 0.7824)
Best baseline R²
0.6708
Effect size d
-0.2908 (95% CI -0.5432 to -0.0391)
Effect size d magnitude
small
Margin over the best yardstick
-0.1105 (95% CI -0.2245 to -0.007842)
Margin's interval stays above zero
No
Model MAE
479.2 (95% CI 384.5 to 580.9)
Model
0.5603 (95% CI 0.3799 to 0.6818)
Model R²
0.5603 (95% CI 0.3799 to 0.6818)
Held-out rows
54
Held-out rows without an answer
0
Target
Wage
Rows where the model is closer
35.19% (95% CI 23.82% to 48.52%)
Rows compared
54
Rows level
0
Regression yardstick trained with gaps filled (too few complete rows)
No
Held-out rows with a gap filled for the regression yardstick
9
Training rows of the regression yardstick
195

How to read this

The table scores the model for Wage against simple yardsticks on the same held-out rows. A yardstick is a predictor simple enough to need no explanation: a constant (the average, the middle value, the most common class or level, the average count), a plain regression on the numeric inputs (a straight line, a logistic regression, a Poisson regression), and for a series over time a forecast from its own past (the last value, the value one season back, the recent trend extended). Which of them appear depends on what kind of column Wage is. They are yardsticks only: the analysis has one model, and it is the last row. A regression yardstick answers every row the model answers: where one of its numeric inputs is empty, it reads the average of that input over the training rows. Nothing is filled in for the model. A yardstick whose row is empty could not be scored, and the summary then says what the margin is measured against.

Read the columns from top to bottom. Each kind of target is scored by the measure it is judged on: R² and MAE for a continuous value, with MASE added for a series (below 1 means a typical miss is smaller than the series' usual step from one row to the next); accuracy and balanced accuracy for a yes/no value, with the Brier score where a probability is predicted; accuracy and macro F1 for a class; weighted kappa and the distance in levels for an ordered level; Poisson deviance and RMSE for a count. When the model clears every yardstick by a wide margin, its structure earns its place. When it is level with the regression yardstick, a plain regression describes the data as well. When it is level with the constant, the inputs carry no usable signal for Wage.

The margin over the best yardstick comes with an interval. When that interval includes zero, the lead is not settled on this many held-out rows. The share of rows on which the model is closer than the best yardstick is counted only on the rows where the two differ; for a yes/no value or a class these are the rows where the two disagree.

5.1.5 Generalization

The model explains held-out rows about as well as training rows (held-out R² 0.56 on 54 rows, gap between training and held-out R² -0.127): it generalizes. The structure the formula encodes carries from the training rows to the held-out rows.

Generalization gap
-0.1271
N test
54
Target
Wage
Test error
479.2 (95% CI 384.5 to 580.9)
Test R²
0.5603 (95% CI 0.3799 to 0.6818)
Train error
0.5964
Train R²
0.4332
Verdict
Generalizing

How to read this

The chart sets the score of the Wage model on the rows it trained on beside its score on the held-out rows it never saw, and the gap between the two bars is the finding. The score is the one that suits the target: R² for a quantity, balanced accuracy for a yes/no target, weighted kappa for ordered levels. The held-out figure carries a bootstrap 95% confidence interval, which says how far it could move with a different held-out draw; the training bar has none, because it is measured on the rows that shaped the model.

The verdict reads from the two figures. Generalizing: the held-out score is strong, or training and held-out agree closely; the structure the formula encodes carries to new rows. Overfit: training is far ahead of held-out and the held-out score is not strong on its own; the formula has absorbed noise from the training rows, and the training figure overstates what to expect on new data. Underfit or noisy: held-out is far ahead of training; this is not strength, it usually means an outlier or a flat region in the training rows that the held-out rows happen to avoid. Insufficient: fewer than 30 held-out rows and no strong held-out score, too little evidence for a negative verdict. A model that predicts the same value for every held-out row gets no verdict: there is nothing that could carry over. In every case the width of the held-out interval says how much weight the gap can bear.

5.1.6 How each term responds to its inputs

The formula decomposes into 2 additive terms: clean separable structure. Each curve below shows one term's response to its free inputs, with the other inputs held at the training median. Steep curves carry the prediction; flat curves are decorative.

Additive
Yes
Rows the search trained on, measured
176
Terms in the formula
2
Notation
0.07 × Agesq × Avgmin + 150
Target
Wage

How to read this

The formula for Wage is split into its additive terms, and each term is traced as one of its inputs moves across the training range while the other inputs stay at their medians. Each curve is therefore a profile of the term at the typical row, not an average over the data: it shows what the term contributes at every value of that input when everything else is ordinary.

A term that depends on several inputs gets one curve per input; a constant term gets none. Terms the formula subtracts are drawn with their sign, so a curve below zero pulls the prediction down. At any fixed row, the terms add up to the model's prediction. Where the search measured it, a term also carries its share of the variation of the model's output on the rows the search trained on; that is a share of what the formula returns, not of Wage itself. The notebook lets the reader trim each range to the central 95% of the input. When the formula has no additive structure, the chapter traces the whole formula as a single term along each of its inputs, and the Sensitivity chapter is the better read for how the inputs combine.

5.1.7 How the formula behaves outside the data

The model diverges beyond the observed range of agesq, avgmin: predictions there are not supported by the formula's shape. Extrapolate with caution.

Bounded at infinity
No
Rows the search trained on, measured
176
Inputs the formula reads
2
N singularities
0
Risky inputs
agesq, avgmin
Target
Wage
Table 5.1.7 · Limits at neg infinity
InputKind
AgesqNegative infinity
AvgminNegative infinity
Table 5.1.7 · Limits at pos infinity
InputKind
AgesqPositive infinity
AvgminPositive infinity

How to read this

The formula for Wage is read as a mathematical object, without data, to state how it behaves beyond the range it was trained on. A black box cannot answer that question; a formula can. For each input the chapter reports the limit as the input grows without bound in either direction (whether the formula diverges, settles to a finite value or oscillates, with the sign tracked exactly, so that a term like minus two times the cube of the input is reported as diverging downward on the right and upward on the left), whether the formula stays finite in every direction, and any singularity with its solved location: a division whose denominator reaches zero, a logarithm or square root whose argument leaves its domain, or a periodic pole.

The inputs along which the formula diverges or meets a singularity are listed as the risky ones; along them the formula's shape gives no support for extrapolation. A formula that stays finite everywhere with no singularity carries no structural hazard when extrapolated; the data still ends where it ends, and a prediction outside the training range remains unsupported by evidence. The plain-language summary at the end states the same facts in one paragraph.

5.1.8 Contributions per term
Additive
Yes
N rows
5

The notebook draws this chapter's figure from each held-out row; the row-by-row detail is there.

How to read this

For up to five representative held-out rows, the chapter draws one bar chart showing how each additive term of the formula builds that row's prediction of Wage. The five rows are the worst prediction (largest absolute residual), the best (smallest), the median miss, the row with the smallest prediction and the row with the largest. When two roles fall on the same row, which is common with few held-out rows, the earlier role keeps it and fewer than five charts appear.

In each chart the bars stack the terms' contributions at the row's input values, terms the formula subtracts sit below zero, and the bars sum to the prediction. The dashed line marks the actual value; the distance between it and the top of the stack is the residual. The bars are evaluated at the same precision as the predictions, so they agree with the Prediction Chart. When the formula has no additive structure, each chart holds one bar for the whole formula, and the Sensitivity chapter is the better read for how individual inputs move it.

5.1.9 What each term is worth

Full-formula R² is 0.56. Each bar below shows how much R² drops when that term is replaced by its mean contribution. Large drop: load-bearing term; near zero: decoration the formula could lose; negative: overfit signal (ablation improves R²).

Additive
Yes
Baseline R²
0.5603
Held-out rows
54

The notebook draws this chapter's figure from each held-out row; the row-by-row detail is there.

How to read this

For each additive term of the formula for Wage, the chart shows how much the held-out R² drops when that term is replaced by its mean contribution, so that it no longer varies from row to row. The bars are sorted from the largest drop down, and the dashed line marks the R² of the full model.

A large drop means the term carries real predictive work; the complexity it adds is paid for. A small drop means the term adds little beyond its average value: it may refine a stronger term, or its curve in the chapter "How each term responds to its inputs" may be nearly flat. A negative drop means the held-out R² rises without the term, which is a sign that the term tracks noise in the training rows. Each figure describes what a term contributes inside the model as discovered; the chart diagnoses the model, it is not a guide to removing terms. When the formula has no additive structure, the chart shows one bar for the whole formula.

5.1.10 What-If
Table 5.1.10 · The prediction at the typical row: every input at its median or most common value
ItemValue
Agesq729
Avgmin25
Term 1 adds1275.8
Term 2 adds150
Prediction1425.8
Additive
Yes
Expected coverage
0.8673
Held-out rows for the interval
54
Rows inside the interval
51
Terms in the formula
2
Residual P5
-961.8
Residual P95
1252.7
Residual std
612.5

How to read this

The chapter shows the model's prediction for Wage at one input combination, with the contribution of each term stacked beneath it. The inputs start at their training medians, so the first view is the prediction for a typical row, and each input can move within the range the training rows cover; the notebook lets the reader change the inputs and watch the prediction and the stack update.

In the stack, a long bar marks the term that dominates at this input combination, often the one whose removal costs most in the chapter "What each term is worth". A short bar marks a term that makes a small adjustment here; the chapter "How each term responds to its inputs" shows whether it stays small everywhere or grows toward the edges of the range. A negative bar marks a term that pulls the prediction down at this point, either because the formula subtracts it or because it evaluates below zero.

The inputs stay inside the training range for a reason: the formula computes a value anywhere, but outside that range no training row supports it. When the formula has no additive structure, the stack holds a single bar for the whole formula and the inputs move the prediction directly.

5.1.11 Sensitivity

The prediction responds to 2 inputs. The formula is additive: each curve below shows the change in the prediction per unit of one input (∂ŷ/∂xᵢ) independently. Flat curves mean a constant linear response, bent curves a non-linear response that depends on the position in the range, and zero crossings a response that reverses sign between regions.

Additive
Yes
Terms in the formula
2

How to read this

For each input, the chart traces the rate at which the predicted Wage changes as that input moves across its training range, with the other inputs held at their medians. The rate is the local slope of the prediction with respect to the input, measured by a small finite difference at each point. A table summarizes each input with the mean, minimum and maximum of the absolute slope, which ranks the inputs by how strongly the prediction responds to them.

A flat curve means the model is close to linear in that input: the prediction responds at the same rate everywhere in the observed range, which is easy to state and says nothing about values beyond that range. A curve that grows or shrinks marks diminishing or accelerating returns, and whether that is plausible is a question for the domain. A curve that crosses zero means the prediction rises with the input in one part of the range and falls in another, which is either a real optimum or a sign that the formula bends around a few rows. A curve with a large magnitude marks an input to which the prediction is highly responsive; the chapter "What each term is worth" says whether the term behind it also carries predictive work.

In the table, the mean absolute slope ranks the inputs by typical responsiveness. A minimum close to the maximum means the responsiveness is nearly constant; a wide gap means it varies strongly across the range. The chapter reads the whole formula, so it applies whether or not the formula splits into terms.

5.1.12 How two inputs act together

The formula couples agesq and avgmin: on 150 of 150 row pairs the mixed difference was not zero, so in the model's output the two act together (measured, not assumed). This describes the formula; it does not say that the data hold no other interaction. The heatmaps below show how the prediction's response to one input depends on another (the others held at the median).

Table 5.1.12 · Input pairs tested inside the formula
InputsIn the model's outputRow pairsPairs with a differenceMean absolute mixed difference
Agesq, AvgminAct together150150207.9
Table 5.1.12 · The prediction across Agesq (columns) and Avgmin (rows), the other inputs at their typical values
Avgmin \ Agesq441666892111713431568
2.889239.2284.7330.4375.9421.6467.1
10.2464.8625.4786.7947.31108.71269.3
17.51690.4966.11243.11518.81795.72071.4
24.819161306.81699.42090.22482.82873.6
32.121141.61647.62155.72661.73169.93675.8
39.431367.21988.32612.13233.13856.94478
Additive
Yes
Inputs in different terms act separately
Yes
Rows the search trained on, measured
176
Terms in the formula
2
Input pairs not tested
0

How to read this

When a single term of the formula combines two or more inputs, the prediction's response to one of them depends on the value of the other. This chapter draws one surface for each such pair: the predicted Wage across the plane of the two inputs, with every other input held at its training median. Pairs that never share a term have no interaction by construction, and the curves in the chapter "How each term responds to its inputs" already say everything about them. Where the search tested a pair, the summary says on how many pairs of rows the two inputs acted together in the model's output; that is a statement about the formula, not a finding that the data hold no other interaction.

Parallel stripes mean the two inputs act almost independently, so the surface is close to a sum of two one-input curves. Curved contours mean a real interaction, where the slope along one input changes with the other; the location of the peak or trough matters. A saddle, rising along one axis and falling along the other, marks a trade-off between the two. A sharp edge or cliff means the formula encodes a change of regime, and the curve for the same input in the chapter "How each term responds to its inputs" shows whether the bend appears there too.

The grid is deliberately coarse, and the color scale runs from low to high predictions. When the formula has no additive structure, the single whole-formula term holds every input, so every pair appears; the number of surfaces is capped for models with many inputs.

5.1.13 Worst Predictions

The worst held-out row was missed by 1,468 (target's scale). The 10 rows with the largest absolute residuals out of 54 held-out rows are shown. Each card breaks the failure down into per-term contributions and input values, so the failure mode is inspectable. A large residual with one concentrated term contribution means one term got this row wrong; a large residual with balanced contributions means a structural mismatch.

Table 5.1.13 · Worst Predictions
ActualPredictedResidualAgesqAvgmin
31251657+146862534.45
35872192+1395108926.79
40002712+128896138.08
29251692+123357638.23
2000867+113310899.4
11202202-108284134.86
1501225-107567622.71
11002152-1052108926.26
1501063-91367619.3
1498626+87257611.81
Additive
Yes
Expected coverage
0.8673
N extrapolated
0
Held-out rows
54
Held-out rows for the interval
54
Rows inside the interval
51
Residual P5
-961.8
Residual P95
1252.7
Residual std
612.5

How to read this

The held-out rows on which the model misses Wage by the most, sorted by absolute residual with the largest first and capped at ten. Each card shows the row's actual value, the prediction, the residual, the contribution of every additive term at that row and the row's input values. R² and MAE say how good the model is on average; these cards say where it fails and let each failure be read input by input and term by term.

A large residual with one dominant term means that term produced the wrong value at this row; the row's value for that input, read against its curve in the chapter "How each term responds to its inputs", shows whether it fell in a region the formula models poorly. A large residual with balanced contributions means every term was a little off and the sum still missed, which points to a shape the formula cannot represent in that region. Residuals that change sign from card to card mean the misses are not a consistent bias but scattered errors on particular rows. When the formula has no additive structure, each card shows a single whole-formula contribution.

5.1.14 Prediction Interval Coverage

Coverage is counted on 54 rows, where one row moves it by 2 points. That is too coarse to judge whether the interval is well calibrated; the table shows the counts.

Table 5.1.14 · Prediction Interval Coverage
Nominal levelHalf-widthEmpirical coverage
50%413.134 of 54 (62.96%)
80%784.943 of 54 (79.63%)
90%1007.446 of 54 (85.19%)
95%1200.450 of 54 (92.59%)
Largest miss (percentage points)
12.96
Held-out rows
54
Rows inside the interval
34, 43, 46, 50
Residual P2.5
-1067.3
Residual P97.5
1360.2
Verdict
insufficient-data

How to read this

Each held-out prediction of Wage carries a prediction interval, and this chapter checks whether that interval covers the actual value as often as it claims. The interval reported with each prediction is built from the held-out residuals themselves: its bounds are the 2.5th and 97.5th percentiles of the residuals added to the prediction, so it is asymmetric when the residuals are skewed and assumes no particular error distribution.

The coverage table checks a second, simpler band: the symmetric interval that a normal error distribution with the residuals' standard deviation would give. At four nominal levels (50%, 80%, 90% and 95%) it lists the half-width of that normal band and the share of held-out rows whose actual value falls inside it. Empirical coverage close to nominal means the residuals are close enough to normal that the two bands agree and the interval can be taken at face value. Empirical coverage below nominal means the normal band is too narrow, the residuals have heavier tails than a normal distribution, and the percentile-based interval is the one to trust. Empirical coverage above nominal means the normal band is wider than the data needs, usually because a few large residuals inflate the standard deviation. Every figure comes from the same held-out residuals the intervals are built from.

5.1.15 Regression Diagnostics

54 held-out residuals, with a spread up to 1,468 in absolute terms. Mean residual is -7.86, close to zero. The model is approximately unbiased on these held-out rows (a small held-out set makes this less reliable; the Prediction Chart chapter states R² and the mean absolute error with their 95% confidence intervals). The charts below check whether the errors look like noise: a pattern in residuals-vs-fitted or a curving Q-Q means a structural miss. The held-out rows are a random draw, so their order carries no time and the autocorrelation of the residuals is not assessed.

Abs residual vs predicted r
0.3606
Bias
-7.861
Bias over residual sd
-0.01284
Held-out rows
54
Order
shuffled
Residual sd
612.5
Share within normal band
0.9259

The notebook draws this chapter's figure from each held-out row; the row-by-row detail is there.

How to read this

Four residual diagnostics for Wage, all computed from the held-out residuals. Each asks in its own way whether the model's errors look like noise, which is what remains when a model has captured all the structure the data holds.

Residuals against predicted values: Each held-out row is plotted with the model's prediction on the horizontal axis and its residual on the vertical axis, around a zero line. A symmetric cloud around zero means errors of similar size across the predicted range and no systematic bias. A funnel or fan means the size of the errors depends on the predicted value: the model is more uncertain in some regions than in others, and its prediction interval is not equally trustworthy everywhere. A curved pattern, a smile or a frown, means the model misses a structural component; the chapters "How each term responds to its inputs" and "How two inputs act together" show the shapes the formula does capture.

Normal Q-Q plot: The sorted residuals are plotted against the quantiles of a normal distribution, with the diagonal marking a perfectly normal set of residuals. Points that hug the line mean approximately normal residuals, so the normal band in the Prediction Interval Coverage chapter will hold. An S-shape means heavier or lighter tails than normal; heavier tails (below the line at the left, above it at the right) mean the model occasionally produces large errors. A banana or arc means skewed residuals: the model is systematically high or low in part of the range.

Histogram: The residuals binned into equal-width intervals. A bell centered on zero means symmetric errors with no bias. An off-center bell means the model is biased, with a mean error away from zero. A long tail or two peaks mean outliers or two regimes inside the prediction range.

Autocorrelation: For each lag, the correlation of the residual sequence with itself shifted by that many rows, with the band that pure noise would stay inside at plus or minus 1.96 divided by the square root of the row count. Bars inside the band mean uncorrelated residuals with no order-dependent structure left. A bar at lag 1 outside the band means the residuals of neighboring dates correlate: structure in time that the model has not captured. A wave across the lags means periodic structure remains, and the Time Series chapter in the Data section is where to look for its period.

Autocorrelation carries meaning only when the row order itself does. When the data has a date column, the held-out rows are the most recent ones in date order, and the autocorrelation is assessed. Without a date column the held-out rows are a random draw: the row after a held-out row is an arbitrary other row, so the autocorrelation is not assessed and the summary above says so.

5.1.16 Search Progress

The search ran for 155 generations, ending at quality 0.354 (gain +0.118 from generation 2). The curve shows a substantial climb. The last improvement came after 98.7% of the search time: the search was still improving when it stopped, so a longer time limit is likely to find a better model.

First generation
2
First quality
23.64%
Last generation
156
Last improvement ms
2,616,027
Last quality
35.4%
Max quality
35.4%
N points
155
N total generations
155
Search time ms
2,650,057
Stop reason
Maximum runtime reached

How to read this

The line traces the best quality the search for the Wage model had reached at each generation, from the first generation to the last. Quality is the technology's combined score on the training rows, the figure the Model Metrics chapter defines. A curve that flattens well before the last generation means the search converged: the budget was more than enough, and the model is the one the search settled on. A curve still rising at the last generation means the budget ended before the search settled, and the model is the best the search reached within it. The chart describes the search, not the data: it says whether the search had time to settle, not whether more rows would help.

5.1.17 Model Constraints

Across 54 held-out predictions, the formula is mathematically well-behaved: every output is finite, stays inside the training-target ±2σ band, and no prediction runs away beyond the bounded-prediction limit. It is safe to consume downstream at face value.

Table 5.1.17 · Model Constraints
NameStatusDetail
Output coherentPassno non-finite predictions across 54 rows
Within domainPassno rows outside the training target's ±2σ range across 54 rows
Bounded predictionsPassno runaway predictions beyond ±10σ across 54 rows
Held-out rows
54
N non finite
0
N outside domain
0
N runaway
0
Verdict
Clean

How to read this

Sanity checks on the Wage formula, run against the held-out rows. Three checks: every prediction is a finite number (output coherent), predictions stay within the range of the training target widened by two standard deviations on each side (within domain), and no single prediction runs away more than ten standard deviations from the training mean (bounded predictions).

These are pass or fail guards, not quality measures: a formula can clear all three and still predict poorly. A failure is the stronger signal. It means the formula misbehaves numerically somewhere in the data it was shown, and any downstream use of it should wait until that is understood.

5.1.18 Modeling

Measured during the search, on the rows the model was trained on

MAE of a straight line through the inputs (same rows)
503.5
RMSE of a straight line through the inputs (same rows)
679.7
Training MAE (rows the search trained on)
602.5
MAE of always predicting the mean (same rows)
826.7
RMSE of always predicting the mean (same rows)
992.1
Training R² (rows the search trained on)
0.4036
Training RMSE (rows the search trained on)
766.2
Rows the search trained on, measured
176
Error
59.64%
Inputs the straight line used
12
N train
215
Quality
32.72%

How to read this

The function below trains the model for Wage on the processed data. Its output is the model that every other chapter of this section scores and dissects.

5.2 Findings

3 findings from the model diagnostics (1 Critical, 2 Info) are listed in the Concerns chapter at the top of the analysis.

6 Implications

The Implications section states how to act on the models: the Stakeholder Summary and the Recommended Next Data, where the models hold and where they weaken (Quality by Input Range, Rows unlike the ones the model learned from, Per-Segment Predictions, Fairness), what to weigh before relying on them (Causal Honesty, Deployment Readiness), and the Provenance & Audit trail a reviewer needs to retrace the analysis.

6.1 Recommended Next Data

  • Collect more rows where Draft is missing. Reason: the analysis warns that Wage differs where Draft is missing, which may signal that important variation is not captured in these cases.
  • Review and address the model’s bias and prediction parity gaps for Allstar, Black, Center and Children. Reason: fairness flags indicate parity and bias for Allstar and prediction parity gaps for Black, Center and Children.
  • Inspect and improve the model for Wage to close the gap with a straight line through the data. Reason: the model explains 56% of the variation in Wage, but loses to a straight line which explains 67%, an 11-point margin that is not settled as the margin’s interval includes zero.
  • Add or review columns that clarify how Lwage relates to Wage. Reason: the analysis finds that Lwage reproduces Wage, suggesting potential redundancy or an unmodeled transformation.

6.2 Quality by Input Range

Quality by Input Range splits the held-out rows by every usable column besides the target and compares the model's quality on each segment with its quality on all held-out rows. A column the model does not read, such as one left out because it leaks the target, is split as well: its segments show where the predictions are weak, not what the model uses. For a number the quality is R², the share of variation explained; for a yes-or-no target it is accuracy, the share of rows classified correctly, the measure the Classification Performance chapter reports. A numeric column is split into quartiles; a categorical column with at most ten values is split by value. A segment with fewer than three held-out rows is left out, because its gap would be noise.

The chart is sorted from the worst gap to the best. Each bar is the segment's figure minus the figure on all held-out rows.

  • Negative bars: the model is weaker on these rows than on all held-out rows. A segment 0.25 or more below the overall value (25 points of accuracy) reads much weaker; a segment 0.10 to 0.25 below reads weaker. These are the segments to treat with caution.
  • Positive bars: the model is stronger on these rows than overall. A strong segment can mask weakness elsewhere, so the negative segments come first.
6.2.1 Wage by Input Range

Segment children = no performs in line with the aggregate (37 held-out rows): segment R² 0.454 against aggregate R² 0.56 (gap -0.106). No quality drift detected.

Table 6.2.1 · Wage by Input Range
SegmentRowsR²Gap to all rowsVerdict
Children = false370.4539-0.1065In line with all rows
Center = false510.5373-0.0230In line with all rows
Black = true480.5371-0.0233In line with all rows
Allstar = false470.5279-0.0324In line with all rows
Guard = false310.6273+0.0669In line with all rows
Marr = false330.5089-0.0514In line with all rows
Marrblck = false360.5582-0.0022In line with all rows
Years played in college = q4430.5507-0.0097In line with all rows
Assists = q4130.2306-0.3297Too few rows to judge
Lwage = q411-1.2218-1.7821Too few rows to judge

The first 10 rows of 24. The other 14 are in the notebook.

Aggregate MAE
479.2
Aggregate R²
0.5603
Measure
r2
Target
Wage

6.3 Rows unlike the ones the model learned from

0 of 54 held-out rows (0%) lie further from the training rows than 99 in 100 rows like them would, measured on the 2 inputs the model reads (agesq, avgmin). That is a small minority; the held-out figures describe the model on rows like the ones it was trained on.

Table 6.3 · Rows unlike the ones the model learned from
TargetInputs measuredLineHeld-out rowsBeyond the lineShare
WageAgesq, Avgmin3.0355400%
Table 6.3 · Score bins · 24 of 30 rows shown
CountFromTo
000.121
10.1210.243
20.2430.364
10.3640.486
00.4860.607
40.6070.728
20.7280.85
40.850.971
40.9711.093
61.0931.214

The first 10 rows of 24. The other 14 are in the notebook.

How to read this

Out-of-distribution detection finds held-out rows that lie far from the training rows of a model. It looks only at the numeric inputs the model reads. Each of them is standardized against the mean and standard deviation of the training rows, and the standardized values are combined into one distance per row. The line a row is read against depends on the number of inputs: it is the distance that 99 in 100 rows like the training rows stay below. A row beyond the line is flagged as out of distribution, and the held-out figures are unlikely to apply to it. When the held-out rows resemble the training rows, about one row in a hundred is flagged.

The chart shows one bar per held-out row in row order; the flagged rows are the bars at or above the line. With more than one target, the chart shows the model with the largest flagged share, and the figures name it.

The check treats inputs independently, so it will not notice a rare combination of values that are each ordinary on their own, such as a large height with a small weight. It is a first screen. A decision that depends on catching unusual inputs needs a dedicated check.

6.4 Fairness

This chapter checks the held-out predictions for uneven treatment of groups. For every categorical column with two to ten distinct values, other than the target, it reports per group the row count, mean actual value, mean prediction and mean residual, with R² and MAE for a number or accuracy, the share of rows classified correctly, for a yes-or-no target, and two measures across the groups.

  • Prediction parity ratio: the smallest group mean prediction divided by the largest, in absolute value. A ratio of 1.0 is perfect parity. Below 0.80, the four-fifths rule, some group receives systematically different predictions.
  • Largest mean residual: the largest absolute mean residual of any group. Above 30% of the overall residual standard deviation, at least one group is systematically over- or under-predicted.

The verdict combines the two: balanced when neither measure trips, parity gap or residual bias when one does, parity gap and residual bias when both do, and insufficient data when fewer than two groups have enough held-out rows.

Diafunc does not decide which columns are protected attributes; it reports every plausible one. A flagged column matters where it is a protected attribute in the decision the model supports, and can be set aside where it is not.

6.4.1 Wage Fairness

Column Allstar splits the held-out rows into 2 groups. The prediction-parity ratio is 0.61 (smallest over largest mean prediction), the largest absolute group residual mean is 391 (0.639 times the overall residual standard deviation). Parity gap and residual bias: the model both predicts systematically lower for some groups and errs systematically against at least one group. This is the worst-case fairness signal on this column; review before deployment. points goes closely with allstar (association 0.607 on a scale from 0 to 1): a model that reads points can tell the groups of allstar apart, so leaving allstar out of the inputs does not keep it out of the predictions.

Aggregate residual std
612.5
N columns
8
N proxies
6
Proxy threshold
0.5
Table 6.4.1 · Columns
Bias max abs residual meanColumnGroupsInterpretationN groupsN proxiesPrediction parity ratioVerdict
391.5Allstartrue, falseColumn Allstar splits the held-out rows into 2 groups. The prediction-parity ratio is 0.61 (smallest over largest mean prediction), the largest absolute group residual mean is 391 (0.639 times the overall residual standard deviation). Parity gap and residual bias: the model both predicts systematically lower for some groups and errs systematically against at least one group. This is the worst-case fairness signal on this column; review before deployment. points goes closely with allstar (association 0.607 on a scale from 0 to 1): a model that reads points can tell the groups of allstar apart, so leaving allstar out of the inputs does not keep it out of the predictions.210.6096Parity and bias
134.1Blacktrue, falseColumn Black splits the held-out rows into 2 groups. The prediction-parity ratio is 0.777 (smallest over largest mean prediction), the largest absolute group residual mean is 134 (0.219 times the overall residual standard deviation). Parity gap: at least one group receives systematically smaller predictions on average. Whether that is a property of the target itself or a model bias against the group needs checking.20.7772Prediction parity gap
13.13Centerfalse, trueColumn Center splits the held-out rows into 2 groups. The prediction-parity ratio is 0.379 (smallest over largest mean prediction), the largest absolute group residual mean is 13.1 (0.0214 times the overall residual standard deviation). Parity gap: at least one group receives systematically smaller predictions on average. Whether that is a property of the target itself or a model bias against the group needs checking.20.3789Prediction parity gap
22.53Childrenfalse, trueColumn Children splits the held-out rows into 2 groups. The prediction-parity ratio is 0.695 (smallest over largest mean prediction), the largest absolute group residual mean is 22.5 (0.0368 times the overall residual standard deviation). Parity gap: at least one group receives systematically smaller predictions on average. Whether that is a property of the target itself or a model bias against the group needs checking.20.6951Prediction parity gap
89.35Forwardtrue, falseColumn Forward splits the held-out rows into 2 groups. The prediction-parity ratio is 0.873 (smallest over largest mean prediction), the largest absolute group residual mean is 89.4 (0.146 times the overall residual standard deviation). Balanced: every group is within the 80% parity band and no group's residual mean exceeds 30% of the overall residual standard deviation. No systematic disparity was detected on this column. guard goes closely with forward (association 0.707 on a scale from 0 to 1): a model that reads guard can tell the groups of forward apart, so leaving forward out of the inputs does not keep it out of the predictions.210.8726Balanced
99.29Guardfalse, trueColumn Guard splits the held-out rows into 2 groups. The prediction-parity ratio is 0.999 (smallest over largest mean prediction), the largest absolute group residual mean is 99.3 (0.162 times the overall residual standard deviation). Balanced: every group is within the 80% parity band and no group's residual mean exceeds 30% of the overall residual standard deviation. No systematic disparity was detected on this column. forward goes closely with guard (association 0.707 on a scale from 0 to 1, and 1 more column at 0.5 or above): a model that reads forward can tell the groups of guard apart, so leaving guard out of the inputs does not keep it out of the predictions.220.9995Balanced
91.12Marrtrue, falseColumn Marr splits the held-out rows into 2 groups. The prediction-parity ratio is 0.885 (smallest over largest mean prediction), the largest absolute group residual mean is 91.1 (0.149 times the overall residual standard deviation). Balanced: every group is within the 80% parity band and no group's residual mean exceeds 30% of the overall residual standard deviation. No systematic disparity was detected on this column. marrblck goes closely with marr (association 0.802 on a scale from 0 to 1): a model that reads marrblck can tell the groups of marr apart, so leaving marr out of the inputs does not keep it out of the predictions.210.8846Balanced
122.4Marrblcktrue, falseColumn Marrblck splits the held-out rows into 2 groups. The prediction-parity ratio is 0.861 (smallest over largest mean prediction), the largest absolute group residual mean is 122 (0.2 times the overall residual standard deviation). Balanced: every group is within the 80% parity band and no group's residual mean exceeds 30% of the overall residual standard deviation. No systematic disparity was detected on this column. marr goes closely with marrblck (association 0.802 on a scale from 0 to 1): a model that reads marr can tell the groups of marrblck apart, so leaving marrblck out of the inputs does not keep it out of the predictions.210.8606Balanced

6.5 Causal Honesty

Diafunc found a formula that captures the structure in the training data. That structure is correlative: it states what predicts what in this data. It does not establish that any input causes the target.

A correlative model is the right tool for predicting the target on new rows like these, for exploring which inputs carry information about the target, and for an early-stage diagnostic or screening tool.

A correlative model is not the right tool for claiming that intervening on an input will change the target, for attributing a target's values to specific causes for accountability, or for choosing between policies that act on the inputs.

The Quality by Input Range, Fairness and Concerns chapters show the most common confounding patterns in the data.

The model is correlative: it captures which columns predict the target in the training data, not whether changing one would change the other. Predictions are reliable for inputs that look like the training distribution; they are not evidence that intervening on an input will change the target. The 4 upgrade pathways below show what evidence would support a causal claim.

Table 6.5 · Upgrade pathways
NameSummary
Controlled experimentRandomize the input to be acted on, then analyze again. Randomization breaks confounding, so any change in the target is attributable to the input.
Natural experimentFind a setting where the input varied for reasons unrelated to the target, an instrumental variable. Variation of that kind supports causal inference.
Causal graphWrite out what plausibly causes what, then check whether the model's terms are consistent with the conditional independences the graph implies.
Control for confoundersIf a third variable plausibly moves both the input and the target, stratify by it or condition on it before training again. Gaps in the Quality by Input Range chapter are a clue.

6.6 Deployment Readiness

The analysis ran with the configuration below. The checks and the verdict of this chapter are read against it.

  • Targets: 1 target: wage
  • Rows: 269 rows
  • Per-Segment Predictions: not produced, because no segment column was chosen
  • Fairness: compared across 8 categorical columns (allstar, black, center, children, forward, guard, marr, marrblck); which of them are protected attributes is not decided here
  • Out-of-Distribution: held-out rows are scored against the training rows; predictions made outside this report are not scored
  • Provenance & Audit: recorded in the chapter of that name

Across 6 checks (3 pass, 2 warn, 1 fails, 1 critical concern), a check fails and a critical concern is open: Not ready to use yet. The failing checks below and the critical concerns in the Concerns chapter need addressing before the predictions are used.

Table 6.6 · Deployment Readiness
NameStatusDetail
GeneralizationPassquality on the held-out rows matches quality on the rows the model was found on
Prediction Interval CoverageWarntoo few held-out rows to judge coverage, largest gap 13.0 percentage points
Baseline Lift for WageFailloses to the best baseline, R² lift -0.11
Model ConstraintsPassevery held-out prediction is finite and within the training range
FairnessWarn8 columns compared, 4 flagged: Allstar (a group both receives smaller predictions and is systematically mis-predicted), Black (a group receives systematically smaller predictions), Center (a group receives systematically smaller predictions), Children (a group receives systematically smaller predictions)
Out-of-DistributionPass0% of held-out rows are out of distribution
N critical concerns
1
N fail
1
N pass
3
N warn
2
Overall verdict
Not ready

6.7 Provenance & Audit

This record captures the inputs and configuration that produced the report, so the analysis can be reproduced or the result defended in an audit.

Table 6.7 · Provenance & Audit
ItemValue
Analysis ID01760490-48b8-11f1-8087-d378640fe787
Run time2026-11-07 14:24 UTC
Total rows269
Modeled targetsWage
Segment columnnot available
Time limit1 h 0 min
Time taken1 h 7 min
Credit budget500 credits
Ufinq version0.3.58
Left out of the inputsLwage

7 Data

The columns are mostly numeric, and some pairings such as age and squared age show a near-perfect correlation. Allstar is categorical and appears relevant for wage. One variable has missing values. Differences in wage where Draft is missing are flagged as a concern; see the Concerns chapter.

Most notable in the Data section:

  • Age and Agesq move together (correlation 0.997)
  • Exper and Expersq move together (correlation 0.95)
  • Agesq and Exper move together (correlation 0.942)

How to read this

The Data section describes the table as supplied, before any modeling: the Column Profiles overview of every column and its role, one chapter per column with its distribution and correlations, a closer look at columns with notable characteristics, Duplicate Rows, Leakage Suspicion, a map for paired coordinate columns, and the General chapter with descriptive statistics, correlations, missing values, outlier ranges, clusters and segments across all columns.

7.1 Column Profiles

Profiled 22 columns across 269 rows (14 numeric, 8 non-numeric): every column is at least 75% populated. Distributional statistics on numeric columns can be read at face value. 7 columns are strongly skewed: the mean of such a column is pulled toward its long tail, so the median is the better guide to a typical value.

Table 7.1 · Columns
NameRoleTypeNMissingMeanStdMinP25MedianP75MaxSkewnessCV
WageTargetnumber2690142410001506501186201557400.96220.7022
AgeInputinteger269027.393.3921252730410.76490.1238
AgesqInputinteger2690761.9195.144162572990016811.1120.2561
AllstarInputbool2690
AssistsInputnumber26902.4092.0930.0000.9001.9003.40012.6001.6880.8688
AvgminInputnumber269023.989.732.8916.7324.8233.2643.09-0.20210.4058
BlackInputbool2690
CenterInputbool2690
ChildrenInputbool2690
Years played in collegeInputinteger26903.7170.75404444-3.5780.2029

The first 10 rows of 22. The other 12 are in the notebook.

N columns
22
N high missing
0
N rows
269
N strongly skewed
7
Table branch
main
Table ID
00bcdd37-a300-4a19-a0a1-93a9c4550e99

How to read this

One row per column of the analyzed table: its role in the analysis (the target being predicted, an input it is predicted from, or the segment column), its type, value count and missing count, and for numeric columns the mean, standard deviation, minimum, quartiles, maximum and interquartile range. A column with 25% or more missing values is flagged; its mean and quartiles describe only the rows that have a value.

Two further figures describe the shape of a numeric column. The skewness says whether the values lean to one side: close to zero the column is symmetric, from 0.5 it is moderately skewed and from 1 strongly skewed, with the longer tail on the high side when the figure is positive and on the low side when it is negative. In a strongly skewed column the mean is pulled toward the tail, and the median is the better guide to a typical value. The coefficient of variation is the standard deviation divided by the mean: it compares how widely columns spread regardless of their units, and is given only for a column whose mean is positive.

These are the columns as supplied, before any modeling: the reference against which every later chapter reads.

7.2 Age

Age: 21 to 41, median 27, 18 distinct values; moves with Agesq most (r 0.997).

How to read this

A column chapter profiles one column: its distribution in figures and charts, and how it relates to the other columns. It opens with one line on what the column holds. The Description chapter carries its figures, with every other column summarized for each of its values where it has few, and its association with every other column where it is numeric. The Visualization chapter carries its charts. A table with a date column adds an Over Time and a Time Series chapter; a target with classes adds a Class Balance chapter.

7.2.1 Description
Table 7.2.1 · Age
StatisticValue
Typeinteger
Values present269
Distinct values18
Continuousyes
Decimal places0
Minimum21
Maximum41
Mean27.39
Median27
Mode24

The first 10 rows of 17. The other 7 are in the notebook.

How to read this

The figures below summarize Age in numbers: where its values center (mean, median, mode), how far they spread (range, variance, standard deviation, interquartile range) and how they are distributed across percentiles. They are the reference for every chart in the Visualization chapter and for the correlation measures further down.

The figures for Age are grouped: the column itself first, then where its values center, how far they spread, the shape of their distribution and its percentiles. Each statistic is explained in the Description chapter under General in the Data section.

7.2.1.1 By Value

Age accounts for 17% of the spread in Wage: 22 averages 1,940 above 25 (95% interval -221 to 4,102). An association, not a cause.

Table 7.2.1.1 · Groups
TargetValueRowsAverage or share95% interval
Wage2232877.7527.4 to 5227.9
Wage32102194.11294 to 3094.2
Wage3542172.51203.7 to 3141.3
Wage33132030.21502.9 to 2557.5
Wage29221895.31346.7 to 2444
Wage31211879.91423.6 to 2336.1
Wage282417721423.3 to 2120.7
Wage30181545.81065 to 2026.6
Wage27211307.2881.3 to 1733.1
Wage26331121802.9 to 1439.2

The first 10 rows of 13. The other 3 are in the notebook.

Table 7.2.1.1 · Mean per value of Age
AgeRowsReboundsYears played in collegeMinutesExperPointsAssistsAvgminDraftGamesAgesqExpersqLwageWage
2119.122438217.53.632.087644147.3781600
2237.9672.3332501.71.66719.572.534.912.33369.6748437.9292877.7
23172.5823.4711229.11.17682.02418.9317.1857.185291.5296.588937.6
24443.8733.8861543.22.0919.252.08422.3318.4564.915765.1826.607998.9
25324.1223.7811346.82.5948.51.82820.219.7761.316258.0316.458937.2
26334.3123.5151493.73.5768.8152.032220.2862.7367614.066.681121
27213.9143.7141576.44.4769.7672.04322.5516.9462.7172921.336.891307.2
28244.9213.7082004.15.41711.883.72528.4920.1368.7578431.57.3311772
29225.0143.9092086.56.513.053.16829.0419.3269.2784142.867.3571895.3
30184.5173.8891802.77.5569.3722.32824.2722.4470.6790058.677.0471545.8

The first 10 rows of 18. The other 8 are in the notebook.

How to read this

The rows are grouped by the value of Age, and every other column is summarized within each group with the same descriptive statistics. A statistic that shifts from group to group marks a column that moves with Age; one that stays level marks a column that does not. Groups with few rows produce unstable figures, so the row count of each group is the first thing to read. Where the groups differ in a target by more than chance would give, the chapter states by how much; it states that the two go together, not that one brings the other about.

7.2.1.2 Correlation
Table 7.2.1.2 · Correlations
ColumnPair ofAssociationPearson's r95% CI (Pearson)Spearman's ρKendall's τq-valueRows
Agesqtwo numbers10.99680.996 to 0.997511< 0.001269
Expertwo numbers0.94120.94120.9258 to 0.95340.91830.8201< 0.001269
Expersqtwo numbers0.91830.90460.8803 to 0.92420.91830.8201< 0.001269
Marra number and a category0.3673< 0.001269
Lwagetwo numbers0.36370.31840.2066 to 0.42190.36370.2566< 0.001269
Wagetwo numbers0.36370.34240.2324 to 0.44390.36370.2566< 0.001269
Marrblcka number and a category0.2441< 0.001269
Childrena number and a category0.19250.003977269
Minutestwo numbers0.18780.14710.02795 to 0.26210.18780.13090.03683269
Avgmintwo numbers0.18710.14110.02184 to 0.25630.18710.13140.04334269

The first 10 rows of 21. The other 11 are in the notebook.

How to read this

Each row pairs Age with one other column, numeric or categorical, and reports how strongly the two go together on an association from 0 to 1, and whether an association this strong could be chance. For another numeric column the row adds the direction: Pearson's coefficient measures a straight-line relationship, and Spearman's rho and Kendall's tau measure a monotonic relationship on ranks, which makes them resistant to outliers and lets them see a curved but consistently rising or falling pattern. For a categorical column the association is the correlation ratio η, which has no direction. A strong association is not an explanation: two columns can go together when a third one moves both.

The table lists every other numeric or categorical column against Age, with the pair's kind and association, three correlation coefficients and a confidence interval where the other column is numeric, the q-value and the rows the pair was measured on.

Table 7.2.1.2 · Correlation
MeasureMeaning
KindWhat the other column holds: a number (numeric) or categories (numeric_categorical). A categorical column is one that is not numeric and holds between 2 and 50 distinct values
AssociationThe strength of the pair from 0 to 1. With a numeric column: the stronger of Pearson's coefficient and Spearman's ρ, without its sign. With a categorical column: the correlation ratio η, the square root of the share of Age's spread that lies between the categories
CovarianceNumeric partner only. The direction of the relationship (positive or negative) in the units of the two columns. It is not standardized, so its size says little about strength without the columns' own spread
Pearson's Correlation CoefficientNumeric partner only. The strength and direction of a linear relationship, from -1 (a perfect falling line) through 0 (no linear relationship) to 1 (a perfect rising line)
Spearman's ρNumeric partner only. The strength and direction of a monotonic relationship, computed on ranks rather than values, from -1 to 1. It follows a curved pattern as long as it rises or falls consistently, and outliers move it little
Kendall's τNumeric partner only. The strength and direction of association from the agreement and disagreement of rank order across pairs of rows, from -1 to 1. It is the steadier of the two rank measures when the rows are few
q-valueHow likely a pair this strong is among unrelated columns, adjusted for the number of pairs involving Age (Benjamini-Hochberg). Among all pairs reported with q below 0.05, about 5% or fewer are expected to be false discoveries
95% CI (Pearson)Numeric partner only. The lower and upper bounds of the 95% confidence interval on the Pearson coefficient, from the Fisher z-transform. An interval that contains 0 leaves the direction of the relationship open
RowsThe rows in which both columns hold a value. The q-value and the interval of a pair rest on this count
7.2.2 Visualization

The rows of Age are not charted in their order: consecutive rows are no more alike than rows picked at random (a correlation of 0.00 between each value and the next), so the order carries no pattern to show.

How to read this

The charts in this chapter were chosen from the type of Age and of the columns it is compared with: a numeric column gets a histogram, a box plot and, when its row order carries a pattern, a value sequence, a categorical column gets a frequency chart, and its relation to each target of the analysis, and to the segment column when there is one, gets the chart that suits the pair, with the target on the vertical axis where it is a number. A column is not charted against every other column; which columns move together is shown under Strongest Correlations. The charts complement the figures in the Description chapter; the shape of a distribution and the position of its outliers are easier to see than to read.

7.3 Agesq

Agesq: 441 to 1,681, median 729, 18 distinct values; moves with Age most (r 0.997).

7.3.1 Description
Table 7.3.1 · Agesq
StatisticValue
Typeinteger
Values present269
Distinct values18
Continuousyes
Decimal places0
Minimum441
Maximum1681
Mean761.9
Median729
Mode576

The first 10 rows of 17. The other 7 are in the notebook.

7.3.1.1 By Value

Agesq accounts for 17% of the spread in Wage: 484 averages 1,940 above 625 (95% interval -221 to 4,102). An association, not a cause.

Table 7.3.1.1 · Groups
TargetValueRowsAverage or share95% interval
Wage48432877.7527.4 to 5227.9
Wage1024102194.11294 to 3094.2
Wage122542172.51203.7 to 3141.3
Wage1089132030.21502.9 to 2557.5
Wage841221895.31346.7 to 2444
Wage961211879.91423.6 to 2336.1
Wage7842417721423.3 to 2120.7
Wage900181545.81065 to 2026.6
Wage729211307.2881.3 to 1733.1
Wage676331121802.9 to 1439.2

The first 10 rows of 13. The other 3 are in the notebook.

Table 7.3.1.1 · Mean per value of Agesq
AgesqRowsReboundsYears played in collegeMinutesExperPointsAssistsAvgminDraftGamesExpersqLwageAgeWage
44119.122438217.53.632.087647.378211600
48437.9672.3332501.71.66719.572.534.912.33369.6737.929222877.7
529172.5823.4711229.11.17682.02418.9317.1857.181.5296.58823937.6
576443.8733.8861543.22.0919.252.08422.3318.4564.915.1826.60724998.9
625324.1223.7811346.82.5948.51.82820.219.7761.318.0316.45825937.2
676334.3123.5151493.73.5768.8152.032220.2862.7314.066.68261121
729213.9143.7141576.44.4769.7672.04322.5516.9462.7121.336.89271307.2
784244.9213.7082004.15.41711.883.72528.4920.1368.7531.57.331281772
841225.0143.9092086.56.513.053.16829.0419.3269.2742.867.357291895.3
900184.5173.8891802.77.5569.3722.32824.2722.4470.6758.677.047301545.8

The first 10 rows of 18. The other 8 are in the notebook.

7.3.1.2 Correlation
Table 7.3.1.2 · Correlations
ColumnPair ofAssociationPearson's r95% CI (Pearson)Spearman's ρKendall's τq-valueRows
Agetwo numbers10.99680.996 to 0.997511< 0.001269
Expertwo numbers0.94210.94210.927 to 0.95420.91830.8201< 0.001269
Expersqtwo numbers0.92690.92690.908 to 0.94210.91830.8201< 0.001269
Lwagetwo numbers0.36370.31140.1992 to 0.41550.36370.2566< 0.001269
Wagetwo numbers0.36370.33930.229 to 0.4410.36370.2566< 0.001269
Marra number and a category0.3574< 0.001269
Marrblcka number and a category0.238< 0.001269
Minutestwo numbers0.18780.13510.01573 to 0.25060.18780.13090.0624269
Avgmintwo numbers0.18710.12860.009168 to 0.24450.18710.13140.07346269
Childrena number and a category0.18390.00646269

The first 10 rows of 21. The other 11 are in the notebook.

7.3.2 Visualization

The rows of Agesq are not charted in their order: consecutive rows are no more alike than rows picked at random (a correlation of -0.01 between each value and the next), so the order carries no pattern to show.

7.4 Allstar

Allstar: 2 distinct values, most often no.

7.4.1 Description
Table 7.4.1 · Allstar
StatisticValue
Typebool
Values present269
Distinct values2
Continuousno
Modefalse
Table 7.4.1 · Rows per value
VariableValueRowsShare
Allstarfalse23888.48%
Allstartrue3111.52%
7.4.1.1 By Value

Allstar accounts for 15% of the spread in Wage: yes averages 1,242 above no (95% interval 815 to 1,669). An association, not a cause.

Table 7.4.1.1 · Groups
TargetValueRowsAverage or share95% interval
Wagetrue312522.52109.5 to 2935.5
Wagefalse2381280.71167.1 to 1394.3
Table 7.4.1.1 · Mean per value of Allstar
AllstarRowsReboundsYears played in collegeMinutesExperPointsAssistsAvgminDraftGamesAgesqExpersqLwageAgeWage
false2384.063.7391545.55.0218.9212.12422.3921.8764.47761.536.746.85927.391280.7
true317.0193.5482731.45.87120.114.59736.198.96875.35764.745.297.67227.422522.5

7.5 Assists

Assists: 0 to 12.6, median 1.90, skewed to high values; moves with Avgmin most (r 0.633).

7.5.1 Description
Table 7.5.1 · Assists
StatisticValue
Typenumber
Values present269
Distinct values72
Continuousyes
Decimal places17
Minimum0
Maximum12.6
Mean2.409
Median1.9
Mode0.4000000059604645

The first 10 rows of 17. The other 7 are in the notebook.

7.5.1.1 Correlation
Table 7.5.1.1 · Correlations
ColumnPair ofAssociationPearson's r95% CI (Pearson)Spearman's ρKendall's τq-valueRows
Avgmintwo numbers0.73990.63250.5549 to 0.69920.73990.5446< 0.001269
Pointstwo numbers0.70430.5390.4483 to 0.61870.70430.5105< 0.001269
Minutestwo numbers0.69560.59480.5116 to 0.6670.69560.5045< 0.001269
Guarda number and a category0.5181< 0.001269
Lwagetwo numbers0.44210.35380.2446 to 0.45420.44210.2995< 0.001269
Wagetwo numbers0.44210.32440.2131 to 0.42740.44210.2995< 0.001269
Gamestwo numbers0.43920.34260.2325 to 0.4440.43920.3131< 0.001269
Allstara number and a category0.378< 0.001269
Forwarda number and a category0.3065< 0.001269
Centera number and a category0.2789< 0.001269

The first 10 rows of 21. The other 11 are in the notebook.

7.5.2 Visualization

The rows of Assists are not charted in their order: consecutive rows are no more alike than rows picked at random (a correlation of 0.03 between each value and the next), so the order carries no pattern to show.

7.6 Avgmin

Avgmin: 2.89 to 43.1, median 24.8; moves with Minutes most (r 0.935).

7.6.1 Description
Table 7.6.1 · Avgmin
StatisticValue
Typenumber
Values present269
Distinct values265
Continuousyes
Decimal places16
Minimum2.889
Maximum43.09
Mean23.98
Median24.82
Mode34.79267883300781

The first 10 rows of 17. The other 7 are in the notebook.

7.6.1.1 Correlation
Table 7.6.1.1 · Correlations
ColumnPair ofAssociationPearson's r95% CI (Pearson)Spearman's ρKendall's τq-valueRows
Minutestwo numbers0.93530.93530.9184 to 0.94870.93340.804< 0.001269
Pointstwo numbers0.91860.8870.8584 to 0.910.91860.7636< 0.001269
Assiststwo numbers0.73990.63250.5549 to 0.69920.73990.5446< 0.001269
Lwagetwo numbers0.66760.64110.5648 to 0.70650.66760.4792< 0.001269
Wagetwo numbers0.66760.62180.5425 to 0.69010.66760.4792< 0.001269
Reboundstwo numbers0.66130.6420.5658 to 0.70730.66130.4835< 0.001269
Gamestwo numbers0.58090.58090.4957 to 0.6550.56960.4093< 0.001269
Allstara number and a category0.4536< 0.001269
Drafttwo numbers0.372-0.2676-0.3813 to -0.1459-0.372-0.2519< 0.001240
Expertwo numbers0.28790.22340.1067 to 0.33410.28790.2023< 0.001269

The first 10 rows of 21. The other 11 are in the notebook.

7.6.2 Visualization

The rows of Avgmin are not charted in their order: consecutive rows are no more alike than rows picked at random (a correlation of 0.02 between each value and the next), so the order carries no pattern to show.

7.7 Black

Black: 2 distinct values, most often yes.

7.7.1 Description
Table 7.7.1 · Black
StatisticValue
Typebool
Values present269
Distinct values2
Continuousno
Modetrue
Table 7.7.1 · Rows per value
VariableValueRowsShare
Blacktrue21780.67%
Blackfalse5219.33%
7.7.1.1 By Value
Table 7.7.1.1 · Mean per value of Black
BlackRowsReboundsYears played in collegeMinutesExperPointsAssistsAvgminDraftGamesAgesqExpersqLwageAgeWage
false523.7153.6731426.95.1548.892.41921.2823.361.73781.638.276.73627.771267.2
true2174.5653.7281743.45.11110.532.40624.6319.4566.68757.237.597.00427.31461.4

7.8 Center

Center: 2 distinct values, most often no.

7.8.1 Description
Table 7.8.1 · Center
StatisticValue
Typebool
Values present269
Distinct values2
Continuousno
Modefalse
Table 7.8.1 · Rows per value
VariableValueRowsShare
Centerfalse22382.9%
Centertrue4617.1%
7.8.1.1 By Value
Table 7.8.1.1 · Mean per value of Center
CenterRowsReboundsYears played in collegeMinutesExperPointsAssistsAvgminDraftGamesAgesqExpersqLwageAgeWage
false2234.2073.7261762.35.01310.562.67424.820.1667.2675435.586.95927.271380
true465.3413.6741293.95.638.5151.12619.9820.458.3800.448.116.92127.981636.3

7.9 Children

Children: 2 distinct values, most often no.

7.9.1 Description
Table 7.9.1 · Children
StatisticValue
Typebool
Values present269
Distinct values2
Continuousno
Modefalse
Table 7.9.1 · Rows per value
VariableValueRowsShare
Childrenfalse17665.43%
Childrentrue9334.57%
7.9.1.1 By Value
Table 7.9.1.1 · Mean per value of Children
ChildrenRowsReboundsYears played in collegeMinutesExperPointsAssistsAvgminDraftGamesAgesqExpersqLwageAgeWage
false1764.1223.7611569.34.6149.4382.11222.6120.9164.61735.932.386.83226.921303.6
true934.9293.6341895.86.07511.672.97126.5818.8867.83811.247.847.18128.291651.3

7.10 Years played in college

Coll: 0 to 4, median 4, skewed to low values, 5 distinct values.

7.10.1 Description
Table 7.10.1 · Years played in college
StatisticValue
Typeinteger
Values present269
Distinct values5
Continuousyes
Decimal places0
Minimum0
Maximum4
Mean3.717
Median4
Mode4

The first 10 rows of 17. The other 7 are in the notebook.

7.10.1.1 By Value
Table 7.10.1.1 · Mean per value of Years played in college
Years played in collegeRowsReboundsMinutesExperPointsAssistsAvgminDraftGamesAgesqExpersqLwageAgeWage
075.2291616.42.71410.662.05722.6726.661.86718.713.866.70526.711576.8
116.2138510.2228.272494847.901222700
246.4752696.83.518.15.07533.975.7579.25566.814.57.25823.751601.9
3375.0241869.2612.072.58125.8212.3368.62785.851.357.26727.761743.2
42204.2241635.75.0959.742.34523.5121.8965.19764.136.786.89727.451356.2
7.10.1.2 Correlation
Table 7.10.1.2 · Correlations
ColumnPair ofAssociationPearson's r95% CI (Pearson)Spearman's ρKendall's τq-valueRows
Drafttwo numbers0.20590.1016-0.02537 to 0.22530.20590.16950.5071240
Pointstwo numbers0.1594-0.121-0.2372 to -0.001467-0.1594-0.12810.5071269
Lwagetwo numbers0.1281-0.0561-0.1745 to 0.06393-0.1281-0.10420.5806269
Wagetwo numbers0.1281-0.1056-0.2224 to 0.01415-0.1281-0.10420.5071269
Reboundstwo numbers0.1166-0.1166-0.233 to 0.003-0.1056-0.084640.5071269
Minutestwo numbers0.1137-0.06811-0.1862 to 0.05191-0.1137-0.091240.5071269
Avgmintwo numbers0.105-0.06316-0.1814 to 0.05687-0.105-0.083780.5286269
Expertwo numbers0.087340.08734-0.0326 to 0.20480.028340.023170.5071269
Allstara number and a category0.081040.5071269
Childrena number and a category0.080190.5071269

The first 10 rows of 21. The other 11 are in the notebook.

7.10.2 Visualization

The rows of Coll are not charted in their order: consecutive rows are no more alike than rows picked at random (a correlation of -0.00 between each value and the next), so the order carries no pattern to show.

7.11 Draft

Draft: 1 to 139, median 14.5, skewed to high values, 11% missing; moves opposite to Lwage most (r -0.402).

7.11.1 Description
Table 7.11.1 · Draft
StatisticValue
Typeinteger
Values present240
Distinct values59
Continuousyes
Decimal places0
Minimum1
Maximum139
Mean20.2
Median14.5
Mode1

The first 10 rows of 17. The other 7 are in the notebook.

7.11.1.1 Correlation
Table 7.11.1.1 · Correlations
ColumnPair ofAssociationPearson's r95% CI (Pearson)Spearman's ρKendall's τq-valueRows
Lwagetwo numbers0.5579-0.402-0.503 to -0.2901-0.5579-0.4185< 0.001240
Wagetwo numbers0.5579-0.3625-0.4677 to -0.2472-0.5579-0.4185< 0.001240
Pointstwo numbers0.424-0.325-0.4337 to -0.2068-0.424-0.2939< 0.001240
Avgmintwo numbers0.372-0.2676-0.3813 to -0.1459-0.372-0.2519< 0.001240
Reboundstwo numbers0.3615-0.3026-0.4134 to -0.183-0.3615-0.2488< 0.001240
Minutestwo numbers0.3129-0.2326-0.3489 to -0.1092-0.3129-0.2115< 0.001240
Allstara number and a category0.2314< 0.001240
Years played in collegetwo numbers0.20590.1016-0.02537 to 0.22530.20590.16950.1882240
Forwarda number and a category0.15320.04602240
Guarda number and a category0.14870.04942240

The first 10 rows of 21. The other 11 are in the notebook.

7.11.2 Visualization

The rows of Draft are not charted in their order: consecutive rows are no more alike than rows picked at random (a correlation of -0.09 between each value and the next), so the order carries no pattern to show.

7.12 Exper

Exper: 1 to 18, median 4, skewed to high values, 15 distinct values; moves with Expersq most (r 0.95).

7.12.1 Description
Table 7.12.1 · Exper
StatisticValue
Typeinteger
Values present269
Distinct values15
Continuousyes
Decimal places0
Minimum1
Maximum18
Mean5.119
Median4
Mode2

The first 10 rows of 17. The other 7 are in the notebook.

7.12.1.1 By Value

Exper accounts for 18% of the spread in Wage: 12 averages 1,569 above 1 (95% interval 765 to 2,372). An association, not a cause.

Table 7.12.1.1 · Groups
TargetValueRowsAverage or share95% interval
Wage1242326.81473 to 3180.5
Wage101021381579.6 to 2696.5
Wage8182070.21445.2 to 2695.2
Wage9162027.51511 to 2543.9
Wage11121888.51285.8 to 2491.2
Wage6241767.31423.3 to 2111.2
Wage1351708849.5 to 2566.5
Wage5261658.41167.7 to 2149
Wage7151361.7941 to 1782.5
Wage4321287.9995.9 to 1579.8

The first 10 rows of 13. The other 3 are in the notebook.

Table 7.12.1.1 · Mean per value of Exper
ExperRowsReboundsYears played in collegeMinutesPointsAssistsAvgminDraftGamesAgesqExpersqLwageAgeWage
1362.9833.3611168.86.9031.63617.921.4857.22576.16.27223.97758.1
2414.3513.7321460.99.6611.8121.3318.764.51591.546.53424.29991.2
3283.9963.71416479.8322.58924.3117.9664.07627.996.73925.041148.7
4324.6383.84415149.051.68421.5718.764.66685166.95826.161287.9
5264.2423.7691821.411.333.12325.6922.2466.81763.2257.12527.581658.4
6244.6373.8332106.512.62.78329.0114.7371.67799.3367.34728.251767.3
7154.6873.81930.310.713.68727.3924.1464.53833.7497.05528.871361.7
8185.6393.7782051.811.282.96128.4923.4471.22924.8647.32530.392070.2
9165.2693.8752135.612.83.08728.7426.9472.69961.6817.495312027.5
10106.283.92140.712.93.5728.6918.775.21012.61007.631.82138

The first 10 rows of 15. The other 5 are in the notebook.

7.12.1.2 Correlation
Table 7.12.1.2 · Correlations
ColumnPair ofAssociationPearson's r95% CI (Pearson)Spearman's ρKendall's τq-valueRows
Expersqtwo numbers10.95030.9373 to 0.960711< 0.001269
Agesqtwo numbers0.94210.94210.927 to 0.95420.91830.8201< 0.001269
Agetwo numbers0.94120.94120.9258 to 0.95340.91830.8201< 0.001269
Lwagetwo numbers0.46920.40620.3013 to 0.50150.46920.3423< 0.001269
Wagetwo numbers0.46920.40920.3045 to 0.50410.46920.3423< 0.001269
Marra number and a category0.3283< 0.001269
Avgmintwo numbers0.28790.22340.1067 to 0.33410.28790.2023< 0.001269
Minutestwo numbers0.27240.21430.09715 to 0.32550.27240.1897< 0.001269
Pointstwo numbers0.24030.18940.07138 to 0.30210.24030.17220.003457269
Marrblcka number and a category0.2319< 0.001269

The first 10 rows of 21. The other 11 are in the notebook.

7.12.2 Visualization

The rows of Exper are not charted in their order: consecutive rows are no more alike than rows picked at random (a correlation of -0.00 between each value and the next), so the order carries no pattern to show.

7.13 Expersq

Expersq: 1 to 324, median 16, skewed to high values, 15 distinct values; moves with Exper most (r 0.95).

7.13.1 Description
Table 7.13.1 · Expersq
StatisticValue
Typeinteger
Values present269
Distinct values15
Continuousyes
Decimal places0
Minimum1
Maximum324
Mean37.72
Median16
Mode4

The first 10 rows of 17. The other 7 are in the notebook.

7.13.1.1 By Value

Expersq accounts for 18% of the spread in Wage: 144 averages 1,569 above 1 (95% interval 765 to 2,372). An association, not a cause.

Table 7.13.1.1 · Groups
TargetValueRowsAverage or share95% interval
Wage14442326.81473 to 3180.5
Wage1001021381579.6 to 2696.5
Wage64182070.21445.2 to 2695.2
Wage81162027.51511 to 2543.9
Wage121121888.51285.8 to 2491.2
Wage36241767.31423.3 to 2111.2
Wage16951708849.5 to 2566.5
Wage25261658.41167.7 to 2149
Wage49151361.7941 to 1782.5
Wage16321287.9995.9 to 1579.8

The first 10 rows of 13. The other 3 are in the notebook.

Table 7.13.1.1 · Mean per value of Expersq
ExpersqRowsReboundsYears played in collegeMinutesExperPointsAssistsAvgminDraftGamesAgesqLwageAgeWage
1362.9833.3611168.86.9031.63617.921.4857.22576.16.27223.97758.1
4414.3513.7321460.929.6611.8121.3318.764.51591.56.53424.29991.2
9283.9963.714164739.8322.58924.3117.9664.07627.96.73925.041148.7
16324.6383.844151449.051.68421.5718.764.666856.95826.161287.9
25264.2423.7691821.4511.333.12325.6922.2466.81763.27.12527.581658.4
36244.6373.8332106.5612.62.78329.0114.7371.67799.37.34728.251767.3
49154.6873.81930.3710.713.68727.3924.1464.53833.77.05528.871361.7
64185.6393.7782051.8811.282.96128.4923.4471.22924.87.32530.392070.2
81165.2693.8752135.6912.83.08728.7426.9472.69961.67.495312027.5
100106.283.92140.71012.93.5728.6918.775.21012.67.631.82138

The first 10 rows of 15. The other 5 are in the notebook.

7.13.1.2 Correlation
Table 7.13.1.2 · Correlations
ColumnPair ofAssociationPearson's r95% CI (Pearson)Spearman's ρKendall's τq-valueRows
Expertwo numbers10.95030.9373 to 0.960711< 0.001269
Agesqtwo numbers0.92690.92690.908 to 0.94210.91830.8201< 0.001269
Agetwo numbers0.91830.90460.8803 to 0.92420.91830.8201< 0.001269
Lwagetwo numbers0.46920.31810.2064 to 0.42170.46920.3423< 0.001269
Wagetwo numbers0.46920.34580.236 to 0.4470.46920.3423< 0.001269
Marra number and a category0.2883< 0.001269
Avgmintwo numbers0.28790.12930.009886 to 0.24510.28790.20230.07611269
Minutestwo numbers0.27240.12780.008282 to 0.24360.27240.18970.07611269
Pointstwo numbers0.24030.12040.000819 to 0.23660.24030.17220.09262269
Assiststwo numbers0.20670.07272-0.0473 to 0.19070.20670.14780.3284269

The first 10 rows of 21. The other 11 are in the notebook.

7.13.2 Visualization

The rows of Expersq are not charted in their order: consecutive rows are no more alike than rows picked at random (a correlation of -0.02 between each value and the next), so the order carries no pattern to show.

7.14 Forward

Forward: 2 distinct values, most often no.

7.14.1 Description
Table 7.14.1 · Forward
StatisticValue
Typebool
Values present269
Distinct values2
Continuousno
Modefalse
Table 7.14.1 · Rows per value
VariableValueRowsShare
Forwardfalse15959.11%
Forwardtrue11040.89%
7.14.1.1 By Value
Table 7.14.1.1 · Mean per value of Forward
ForwardRowsReboundsYears played in collegeMinutesExperPointsAssistsAvgminDraftGamesAgesqExpersqLwageAgeWage
false1593.5353.74816585.11910.222.94223.6722.5665.43761.638.186.88127.381381.4
true1105.6533.6731717.15.11810.21.63924.4216.7266.15762.437.067.05627.421485.1

7.15 Games

Games: 3 to 82, median 74, skewed to low values; moves with Minutes most (r 0.788).

7.15.1 Description
Table 7.15.1 · Games
StatisticValue
Typeinteger
Values present269
Distinct values58
Continuousyes
Decimal places0
Minimum3
Maximum82
Mean65.72
Median74
Mode82

The first 10 rows of 17. The other 7 are in the notebook.

7.15.1.1 Correlation
Table 7.15.1.1 · Correlations
ColumnPair ofAssociationPearson's r95% CI (Pearson)Spearman's ρKendall's τq-valueRows
Minutestwo numbers0.78780.78780.7377 to 0.82930.78710.6097< 0.001269
Avgmintwo numbers0.58090.58090.4957 to 0.6550.56960.4093< 0.001269
Pointstwo numbers0.52760.49680.401 to 0.58180.52760.3765< 0.001269
Assiststwo numbers0.43920.34260.2325 to 0.4440.43920.3131< 0.001269
Reboundstwo numbers0.3890.34380.2338 to 0.44510.3890.2724< 0.001269
Lwagetwo numbers0.36990.36990.2619 to 0.46870.35890.244< 0.001269
Wagetwo numbers0.35890.30380.1911 to 0.40850.35890.244< 0.001269
Expertwo numbers0.20370.14820.02912 to 0.26310.20370.14850.03146269
Expersqtwo numbers0.20370.09346-0.02644 to 0.21070.20370.14850.1657269
Allstara number and a category0.18470.006178269

The first 10 rows of 21. The other 11 are in the notebook.

7.15.2 Visualization

The rows of Games are not charted in their order: consecutive rows are no more alike than rows picked at random (a correlation of 0.01 between each value and the next), so the order carries no pattern to show.

7.16 Guard

Guard: 2 distinct values, most often no.

7.16.1 Description
Table 7.16.1 · Guard
StatisticValue
Typebool
Values present269
Distinct values2
Continuousno
Modefalse
Table 7.16.1 · Rows per value
VariableValueRowsShare
Guardfalse15657.99%
Guardtrue11342.01%
7.16.1.1 By Value
Table 7.16.1.1 · Mean per value of Guard
GuardRowsReboundsYears played in collegeMinutesExperPointsAssistsAvgminDraftGamesAgesqExpersqLwageAgeWage
false1565.5613.6731592.35.2699.7011.48823.1117.8563.83773.640.327.01627.581529.7
true1132.83.7791806.24.91210.913.68125.1823.4968.34745.834.136.86427.131277.7

7.17 Lwage

The model does not use this column: it is left out of the model's inputs. Its figures and charts describe the data all the same, and its relation to the target shows what the model is not told.

Lwage: 5.01 to 8.66, median 7.08; moves with Wage most (r 0.894).

7.17.1 Description
Table 7.17.1 · Lwage
StatisticValue
Typenumber
Values present269
Distinct values180
Continuousyes
Decimal places16
Minimum5.011
Maximum8.655
Mean6.952
Median7.078
Mode5.010634899139404

The first 10 rows of 17. The other 7 are in the notebook.

7.17.1.1 Correlation
Table 7.17.1.1 · Correlations
ColumnPair ofAssociationPearson's r95% CI (Pearson)Spearman's ρKendall's τq-valueRows
Wagetwo numbers10.89390.867 to 0.915611< 0.001269
Avgmintwo numbers0.66760.64110.5648 to 0.70650.66760.4792< 0.001269
Pointstwo numbers0.66040.61960.54 to 0.68820.66040.4754< 0.001269
Minutestwo numbers0.60470.59190.5083 to 0.66450.60470.4295< 0.001269
Drafttwo numbers0.5579-0.402-0.503 to -0.2901-0.5579-0.4185< 0.001240
Reboundstwo numbers0.52180.48810.3913 to 0.57420.52180.368< 0.001269
Expertwo numbers0.46920.40620.3013 to 0.50150.46920.3423< 0.001269
Expersqtwo numbers0.46920.31810.2064 to 0.42170.46920.3423< 0.001269
Assiststwo numbers0.44210.35380.2446 to 0.45420.44210.2995< 0.001269
Gamestwo numbers0.36990.36990.2619 to 0.46870.35890.244< 0.001269

The first 10 rows of 21. The other 11 are in the notebook.

7.17.2 Visualization

The rows of Lwage are not charted in their order: consecutive rows are no more alike than rows picked at random (a correlation of -0.04 between each value and the next), so the order carries no pattern to show.

7.18 Marr

Marr: 2 distinct values, most often no.

7.18.1 Description
Table 7.18.1 · Marr
StatisticValue
Typebool
Values present269
Distinct values2
Continuousno
Modefalse
Table 7.18.1 · Rows per value
VariableValueRowsShare
Marrfalse15055.76%
Marrtrue11944.24%
7.18.1.1 By Value
Table 7.18.1.1 · Mean per value of Marr
MarrRowsReboundsYears played in collegeMinutesExperPointsAssistsAvgminDraftGamesAgesqExpersqLwageAgeWage
false1504.4813.7471598.74.1279.5542.11723.0319.7564.57699.925.796.83626.291283.3
true1194.33.6811787.46.3711.042.77725.1820.7667.18840.152.767.09828.791601

7.19 Marrblck

Marrblck: 2 distinct values, most often no.

7.19.1 Description
Table 7.19.1 · Marrblck
StatisticValue
Typebool
Values present269
Distinct values2
Continuousno
Modefalse
Table 7.19.1 · Rows per value
VariableValueRowsShare
Marrblckfalse17866.17%
Marrblcktrue9133.83%
7.19.1.1 By Value
Table 7.19.1.1 · Mean per value of Marrblck
MarrblckRowsReboundsYears played in collegeMinutesExperPointsAssistsAvgminDraftGamesAgesqExpersqLwageAgeWage
false1784.4253.7131608.84.5569.652.24423.1420.6264.61728.730.966.87126.81315.8
true914.3553.7251825.86.2211.312.73125.6319.3767.9826.750.957.11228.551635.2

7.20 Minutes

Minutes: 33 to 3,533, median 1,690; moves with Avgmin most (r 0.935).

7.20.1 Description
Table 7.20.1 · Minutes
StatisticValue
Typeinteger
Values present269
Distinct values254
Continuousyes
Decimal places0
Minimum33
Maximum3533
Mean1682.2
Median1690
Mode877

The first 10 rows of 17. The other 7 are in the notebook.

7.20.1.1 Correlation
Table 7.20.1.1 · Correlations
ColumnPair ofAssociationPearson's r95% CI (Pearson)Spearman's ρKendall's τq-valueRows
Avgmintwo numbers0.93530.93530.9184 to 0.94870.93340.804< 0.001269
Pointstwo numbers0.86490.83760.798 to 0.87010.86490.6855< 0.001269
Gamestwo numbers0.78780.78780.7377 to 0.82930.78710.6097< 0.001269
Assiststwo numbers0.69560.59480.5116 to 0.6670.69560.5045< 0.001269
Reboundstwo numbers0.61430.59560.5125 to 0.66760.61430.4404< 0.001269
Lwagetwo numbers0.60470.59190.5083 to 0.66450.60470.4295< 0.001269
Wagetwo numbers0.60470.56340.4758 to 0.63990.60470.4295< 0.001269
Allstara number and a category0.4247< 0.001269
Drafttwo numbers0.3129-0.2326-0.3489 to -0.1092-0.3129-0.2115< 0.001240
Expertwo numbers0.27240.21430.09715 to 0.32550.27240.1897< 0.001269

The first 10 rows of 21. The other 11 are in the notebook.

7.20.2 Visualization

The rows of Minutes are not charted in their order: consecutive rows are no more alike than rows picked at random (a correlation of 0.03 between each value and the next), so the order carries no pattern to show.

7.21 Points

Points: 1.20 to 29.8, median 9.30; moves with Avgmin most (r 0.887).

7.21.1 Description
Table 7.21.1 · Points
StatisticValue
Typenumber
Values present269
Distinct values159
Continuousyes
Decimal places16
Minimum1.2
Maximum29.8
Mean10.21
Median9.3
Mode6.599999904632568

The first 10 rows of 17. The other 7 are in the notebook.

7.21.1.1 Correlation
Table 7.21.1.1 · Correlations
ColumnPair ofAssociationPearson's r95% CI (Pearson)Spearman's ρKendall's τq-valueRows
Avgmintwo numbers0.91860.8870.8584 to 0.910.91860.7636< 0.001269
Minutestwo numbers0.86490.83760.798 to 0.87010.86490.6855< 0.001269
Assiststwo numbers0.70430.5390.4483 to 0.61870.70430.5105< 0.001269
Lwagetwo numbers0.66040.61960.54 to 0.68820.66040.4754< 0.001269
Wagetwo numbers0.66040.65710.5833 to 0.72010.66040.4754< 0.001269
Allstara number and a category0.6066< 0.001269
Reboundstwo numbers0.57950.56360.4761 to 0.640.57950.4189< 0.001269
Gamestwo numbers0.52760.49680.401 to 0.58180.52760.3765< 0.001269
Drafttwo numbers0.424-0.325-0.4337 to -0.2068-0.424-0.2939< 0.001240
Expertwo numbers0.24030.18940.07138 to 0.30210.24030.17220.003803269

The first 10 rows of 21. The other 11 are in the notebook.

7.21.2 Visualization

The rows of Points are not charted in their order: consecutive rows are no more alike than rows picked at random (a correlation of 0.03 between each value and the next), so the order carries no pattern to show.

7.22 Rebounds

Rebounds: 0.5 to 17.3, median 3.80, skewed to high values; moves with Avgmin most (r 0.642).

7.22.1 Description
Table 7.22.1 · Rebounds
StatisticValue
Typenumber
Values present269
Distinct values91
Continuousyes
Decimal places16
Minimum0.5
Maximum17.3
Mean4.401
Median3.8
Mode2.700000047683716

The first 10 rows of 17. The other 7 are in the notebook.

7.22.1.1 Correlation
Table 7.22.1.1 · Correlations
ColumnPair ofAssociationPearson's r95% CI (Pearson)Spearman's ρKendall's τq-valueRows
Avgmintwo numbers0.66130.6420.5658 to 0.70730.66130.4835< 0.001269
Minutestwo numbers0.61430.59560.5125 to 0.66760.61430.4404< 0.001269
Pointstwo numbers0.57950.56360.4761 to 0.640.57950.4189< 0.001269
Wagetwo numbers0.53980.53980.4492 to 0.61940.52180.368< 0.001269
Lwagetwo numbers0.52180.48810.3913 to 0.57420.52180.368< 0.001269
Guarda number and a category0.472< 0.001269
Gamestwo numbers0.3890.34380.2338 to 0.44510.3890.2724< 0.001269
Drafttwo numbers0.3615-0.3026-0.4134 to -0.183-0.3615-0.2488< 0.001240
Forwarda number and a category0.3606< 0.001269
Allstara number and a category0.3273< 0.001269

The first 10 rows of 21. The other 11 are in the notebook.

7.22.2 Visualization

The rows of Rebounds are not charted in their order: consecutive rows are no more alike than rows picked at random (a correlation of 0.04 between each value and the next), so the order carries no pattern to show.

7.23 Wage

Wage: 150 to 5,740, median 1,186, skewed to high values; moves with Lwage most (r 0.894).

7.23.1 Description
Table 7.23.1 · Wage
StatisticValue
Typenumber
Values present269
Distinct values180
Continuousyes
Decimal places13
Minimum150
Maximum5740
Mean1423.8
Median1186
Mode150.0

The first 10 rows of 17. The other 7 are in the notebook.

7.23.1.1 Correlation
Table 7.23.1.1 · Correlations
ColumnPair ofAssociationPearson's r95% CI (Pearson)Spearman's ρKendall's τq-valueRows
Lwagetwo numbers10.89390.867 to 0.915611< 0.001269
Avgmintwo numbers0.66760.62180.5425 to 0.69010.66760.4792< 0.001269
Pointstwo numbers0.66040.65710.5833 to 0.72010.66040.4754< 0.001269
Minutestwo numbers0.60470.56340.4758 to 0.63990.60470.4295< 0.001269
Drafttwo numbers0.5579-0.3625-0.4677 to -0.2472-0.5579-0.4185< 0.001240
Reboundstwo numbers0.53980.53980.4492 to 0.61940.52180.368< 0.001269
Expertwo numbers0.46920.40920.3045 to 0.50410.46920.3423< 0.001269
Expersqtwo numbers0.46920.34580.236 to 0.4470.46920.3423< 0.001269
Assiststwo numbers0.44210.32440.2131 to 0.42740.44210.2995< 0.001269
Allstara number and a category0.3973< 0.001269

The first 10 rows of 21. The other 11 are in the notebook.

7.23.2 Visualization

The rows of Wage are not charted in their order: consecutive rows are no more alike than rows picked at random (a correlation of -0.11 between each value and the next), so the order carries no pattern to show.

7.24 Duplicate Rows

Across 269 rows the table has no exact duplicates: every row's value vector is unique across all columns, so the row count reflects distinct rows.

Largest group size
0
N exact duplicates
0
N exact groups
0
N rows
269
N unique
269
Pct exact duplicates
0
Verdict
Clean

How to read this

This chapter counts rows that repeat another row exactly, value for value: how many rows take part in a duplicate group, how many groups exist, and how large the largest group is.

Duplicate rows inflate the apparent row count and can place identical rows on both sides of the training and held-out split, which flatters every quality figure in this report. Only exact repeats are detected here; rows that differ by a rounding digit or a whitespace variant count as distinct.

7.25 Leakage Suspicion

Nothing to check. No column that awaits an answer tracks a target at 0.95 or more, so no column on its own gives a target away.

Categorical verdict
Clean
Critical threshold
0.999
N categorical critical
0
N categorical suspects
0
N confirmed
0
N critical
0
N suspects
0
Threshold
0.95
Verdict
Clean

How to read this

For every column that is not a target, this chapter reads how closely the column tracks each target, on an association from 0 to 1, and lists any pair at 0.95 or more. For two numeric columns the association is the stronger of Pearson's r and Spearman's rho; when the column or the target is categorical it is the correlation ratio η, or Cramér's V when both are. At 0.999 or more the column reproduces the target: that is what a copy or a rescaled copy of it looks like.

A listed column poses one question: is its value known before the target's value exists? If it is, the column is a strong input and the quality stated in this report is real. If it was computed from the target or recorded after it, a formula that reads it looks right on these rows and will not hold on new data. The coefficients read the same in both cases; the answer needs someone who knows where the column came from.

The owner of the analysis can give that answer. A column confirmed as known before the target is listed as confirmed and is no longer a suspect. A column that turns out to be a leak is left out of the model's inputs, and the analysis run again.

7.26 General

Clustering found no clear groups. The best grouping of the 240 complete rows over 13 numeric columns is into 3 groups, with a silhouette of 0.246 (on a scale where 1 is fully separate groups and 0 is none), below the 0.50 that clear structure shows, so no Clusters chapter is written.

Table 7.26 · Groupings tried
GroupsSilhouetteInertiaKept
20.24452286.8
30.24641842yes
40.16821680.9
50.19951532.9
60.20011443.2
Inertia
1842
K
3
N columns used
13
N rows
269
N rows complete
240
Silhouette
0.2464
Verdict
Weak

How to read this

The General chapter covers what holds across all columns together: descriptive statistics, correlations, missing values and outlier ranges under Description, the Clusters found among the rows, and, when a segment column was chosen, the Segments.

7.26.1 Description
Table 7.26.1 · Columns
VariableTypeValues presentDistinct valuesContinuousDecimal places
Reboundsnumber26991yes16
Years played in collegeinteger2695yes0
Minutesinteger269254yes0
Forwardbool2692no
Marrbool2692no
Guardbool2692no
Centerbool2692no
Blackbool2692no
Experinteger26915yes0
Pointsnumber269159yes16

The first 10 rows of 22. The other 12 are in the notebook.

Table 7.26.1 · Center
VariableMeanMedianMode
Rebounds4.4013.82.700000047683716
Years played in college3.71744
Minutes1682.21690877
Forwardfalse
Marrfalse
Guardfalse
Centerfalse
Blacktrue
Exper5.11942
Points10.219.36.599999904632568

The first 10 rows of 22. The other 12 are in the notebook.

Table 7.26.1 · Spread
VariableMinimumMaximumStandard deviationVarianceInterquartile rangeSum
Rebounds0.517.32.8938.3673.21183.9
Years played in college040.75440.569101000
Minutes333533893.3798,0351455452,510
Forward
Marr
Guard
Center
Black
Exper1183.411.5651377
Points1.229.85.90134.828.82746.6

The first 10 rows of 22. The other 12 are in the notebook.

Table 7.26.1 · Shape
VariableSkewnessKurtosisTotal sum of squares
Rebounds1.2631.6262242.3
Years played in college-3.57813.9152.5
Minutes-0.1179-1.087213,873,248
Forward
Marr
Guard
Center
Black
Exper0.82060.26883098.2
Points0.66750.072849331.2

The first 10 rows of 22. The other 12 are in the notebook.

Table 7.26.1 · Rows per value
VariableValueRowsShare
Forwardfalse15959.11%
Forwardtrue11040.89%
Marrfalse15055.76%
Marrtrue11944.24%
Guardfalse15657.99%
Guardtrue11342.01%
Centerfalse22382.9%
Centertrue4617.1%
Blacktrue21780.67%
Blackfalse5219.33%

The first 10 rows of 16. The other 6 are in the notebook.

How to read this

This chapter summarizes every column at once: the descriptive statistics table below, then Correlation, Missing Values and Outliers. For a numeric column the table reports its center (mean, median, mode), its spread (standard deviation, variance, interquartile range, percentiles) and its shape (skewness, kurtosis); for a categorical column it reports the count and share of each category.

Each row of the tables describes one column. The statistics are grouped: the column itself, where its values center, how far they spread, what shape their distribution takes, and its percentiles. They read as follows.

Table 7.26.1 · Description
StatisticExplanation
VariableName of the column the statistics in the row describe
TypeType assigned during data processing (bool, integer, number, decimal, text or date)
Values presentNumber of rows in which the column holds a value
Distinct valuesNumber of different values the column holds
ContinuousWhether the column can take any value in a range, as opposed to a discrete column limited to a fixed set of values
Decimal placesLength of the fractional part of the values
MeanArithmetic average: the sum of all values divided by the number of values
MedianThe middle value: half of the values lie at or below it and half at or above it. It is also the 50% percentile
ModeThe value that appears most often
MinimumSmallest numeric value in the column

The first 10 rows of 20. The other 10 are in the notebook.

7.26.1.1 Missing Values

1 column has missing values. 240 of 269 rows (89%) hold a value in every column a model reads; a model is computed on those rows only. draft is the only gap in 29 rows: without it 269 rows would be complete.

N columns with missing
1
N complete rows
240
N rows
269
Pct complete rows
89.22

How to read this

The table lists, for each column, the count of present values (Values) beside the row count of the processed table (Rows), the number of missing values and their share. The larger the share, the less any statistic or model built on the column can be trusted.

  • Below 5% missing: safe to use as it is.
  • 5% to 25% missing: usable; filling the gaps or excluding those rows is worth considering.
  • 25% or more missing: the column is flagged and receives its own chapter in the Data section. Verify why so much is missing before trusting any result that depends on it.

A model is computed only on rows that hold a value in every column it reads, so gaps spread over several columns cost more rows than any one column's share suggests. The last column, Complete rows without, shows for each input with gaps how many rows would be complete if that column were left out; it is blank for a column without gaps, for a target, and for a column no model reads. A column that is the only gap in many rows is the one whose removal keeps the most data.

7.26.1.2 Outliers

10 of 13 numeric columns have values beyond the Tukey fences (1.5 times the interquartile range outside the quartiles). expersq has the largest share: 8.6% of its values. Whether such values are rare but real rows or recording errors decides whether they are kept. 1 column was not assessed: too few values, or a middle half that sits on a single value.

Table 7.26.1.2 · Outliers
ColumnValuesLower fenceUpper fenceBelowAboveShare
Age269none below37.50020.74%
Agesq269none below1312.5020.74%
Assists269none below7.1500103.72%
Avgmin269none belownone above000%
Years played in college269
Draft240none below60.13093.75%
Exper269none below14.500020.74%
Expersq269none below116.500238.55%
Games26924none above1304.83%
Lwage269none belownone above000%

The first 10 rows of 14. The other 4 are in the notebook.

N columns
14
N columns assessed
13
N columns with outliers
10
Worst column
expersq
Worst pct
8.55

How to read this

The table counts the outliers of every numeric column by the Tukey rule. Q1 and Q3 are the values below which 25% and 75% of a column's values fall, and the interquartile range is Q3 minus Q1. The lower fence lies 1.5 times the interquartile range below Q1, the upper fence the same distance above Q3; a value beyond a fence is an outlier. For each column the table gives both fences, how many values lie below and above them, their share of the column's values, and the single value that lies furthest beyond its fence. These are the same fences the box plots draw their whiskers to.

A column is not assessed when it has fewer than four values, or when the middle half of its values sits on a single number: the interquartile range is then zero and the rule would call every other value an outlier.

Extreme values can be real (rare but legitimate rows) or errors (measurement mistakes, data-entry slips). Which case applies decides whether to keep, clip or drop them before trusting a model that uses the column. A strongly skewed column has many values beyond one fence without any of them being wrong.

7.26.1.3 Correlation

231 column pairs analyzed across 269 rows (91 of two numbers, 112 of a number and a category, 28 of two categories); 106 (46%) are unlikely to be chance, after allowing for how many pairs were tested. Any other pair may still show a large association, but one that cannot be told apart from chance on the rows the pair was measured on. Every pair is stated as an association from 0 to 1, whatever its kind, so pairs of different kinds can be ranked against each other. An association states that two columns go together, not that one brings the other about.

N categorical pairs
28
N mixed pairs
112
N numeric pairs
91
N pairs
231
N samples
269
N significant
106

This section continues in the notebook.

The report continues in the notebook.

Loading Diafunc…