# Mid-Atlantic Wages

Analysis report, automatically produced by Diafunc.

- **Target:** Wage
- **Analyzed:** 2026-10-24 17:39 UTC
- **Data:** [R package ISLR (CRAN)](https://cran.r-project.org/package=ISLR), published under https://diafunc.com/licences/public-domain

## Summary

The model answers how Wage moves with Age in this data, describing only a small part of the variation, and only just ahead of a straight-line trend. It tracks Age alone, when Age rises, the model's output rises as well, but this is a slim advance, not enough to show consistent gains row by row. The margin over a straight line is slight and not yet settled as a clear lead.

To rely on the model’s output for individual predictions or decisions is not advised. Its answer captures only a minor pattern, and it is far less accurate for every quarter of Logwage and several Education groups. Results differ between groups for Education, Maritl, and Race. Validate on a fresh sample before using the model for decisions.

## Formula

```
Wage = bitor(80, Age)
```

1 term · single non-separable expression

## Held-out diagnostics

| Target | Quality | Error | R² | MAE | Generalization | Rows |
| --- | --- | --- | --- | --- | --- | --- |
| Wage | 7.38 % | 92.07 % | 0.0793 | 28.93 | generalizing | 600 |

- Wage · Quality 95 % CI 3.14 % – 11.57 %
- Wage · Error 95 % CI 87.91 % – 96.29 %
- Wage · R² 95 % CI 0.03715 – 0.1209
- Wage · MAE 95 % CI 26.67 – 31.52

## Concerns

**CRITICAL** (implications) — Weaker quality on 10 slices of the data

On 10 of the 32 slices compared the model is weaker than on all held-out rows (R² 0.0793). The largest gaps: `Logwage` in the second quarter (4.45 to 4.64) (worse than always guessing the segment's average, on 126 rows) and `Logwage` in the lowest quarter (3.30 to 4.44) (worse than always guessing the segment's average, on 152 rows). Predictions for these rows deserve caution; the Quality by Input Range chapter lists every slice.

**Complete list (10)**

| Column | Detail |
| --- | --- |
| Logwage in the second quarter (4.45 to 4.64) | 126 |
| Logwage in the lowest quarter (3.30 to 4.44) | 152 |
| Logwage in the third quarter (4.65 to 4.85) | 155 |
| Education = 1_HS_GRAD | 53 |
| Logwage in the highest quarter (4.86 to 5.75) | 167 |
| Education = 5_ADVANCED_DEGREE | 86 |
| Maritl = 4_DIVORCED | 36 |
| Education = 2_HS_GRAD | 190 |
| Race = 3_ASIAN | 30 |
| Age in the highest quarter (51 to 73) | 138 |

**WARNING** (model) — The model barely beats a straight line through the data on `Wage`

The model explains 7.9% of the variation on the held-out rows, where a straight line through the data explains 3.8%, a margin of 4.1 points. The formula adds little over the reference predictor; whether it is worth its complexity depends on the use. Row by row the two are about level.

**INFO** (data) — Column `Region` holds almost one value

`Region` is "2_MIDDLE_ATLANTIC" in every one of its 3,000 values. A column this close to constant tells rows apart in a few cases only: a term that reads it rests on those few rows, and its charts show one bar.

**INFO** (model) — `Logwage` reproduces `Wage`

Spearman's rho = 1 (Pearson r = 0.951) between `Logwage` 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 `Logwage` 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 2026-10-24 17:40 UTC, analysis `0a807890-48b8-11f1-8087-d378640fe787`, 1 target, time budget 1 h 0 min, compute budget 500 credits, 3,000 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 very high quality and still tell you nothing new.

## What it found

Wage is highest at a middle value of Age.

1. **Age is the only input the formula reads.** Everything the formula explains about Wage comes from Age.
2. **Wage is highest at a middle value of Age.** With the other inputs at typical values, the formula gives 126 at Age 46, against 82 at 18 and 80 at 80.
3. **Education is not in the formula.** It has the second-strongest association with Wage in the table (0.48).
4. **The formula is only slightly ahead of a straight line.** The formula explains 7.9% of Wage; a straight line explains 3.8%, so treat them as level. That was measured on 600 rows it never saw. The formula typically misses Wage by 28.9. It is weakest where Logwage is between 4.45 and 4.64.

## Summary

**What to do next:** Intervals are wider than their level requires, largest gap 12.2 percentage points.

5 chapters, each headed by what it found.

### Age carries the whole formula

*What the formula relies on.*

| Input | Share of importance | Permutation importance | Sensitivity | Residual correlation | Occurrences | Model's output as the input rises |
| --- | --- | --- | --- | --- | --- | --- |
| Age | 100 % | 0.1649 | 12.43 | -0.02515 | 1 | Rises (share of rows with a rising slope) |

The formula is **not additively decomposable**: the prediction comes from a single non-separable expression. The curve below shows the whole-formula response.

- **Additive:** No
- **Terms in the formula:** 1
- **Notation:** bitor(80, Age)
- **Target:** Wage

### It holds on new rows: R² 0.08 held out, 0.07 on the training rows

*How well it holds.* The formula explains 7.9% of Wage; a straight line explains 3.8%, so treat them as level.

R² of 0.0793: **weak**. 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 28.9 on the target's scale.

- **Error:** 92.07 % (95 % CI 87.91 % – 96.29 %)
- **MAE:** 28.93 (95 % CI 26.67 – 31.52)
- **N:** 600
- **Training rows:** 2400
- **Quality:** 7.38 % (95 % CI 3.14 % – 11.57 %)
- **R²:** 0.0793 (95 % CI 0.03715 – 0.1209)
- **Target:** Wage

The model beats the strongest simple yardstick, a straight line through the data, by 4.1 points of variance explained on the held-out rows. Row by row the margin is **negligible**: the model reads as roughly level with the yardstick. Of the 600 held-out rows where the two differ, the model is closer on 53% (95% interval 49% to 57%). The interval of the margin includes zero: on this many held-out rows the lead is not settled.

- **Best yardstick's score:** 0.03834 (95 % CI 0.003683 – 0.06956)
- **Best baseline R²:** 0.03834
- **Effect size d:** 0.05844 (95 % CI -0.02002 – 0.1411)
- **Effect size d magnitude:** negligible
- **Margin over the best yardstick:** 0.04096 (95 % CI -0.003399 – 0.0816)
- **Margin's interval stays above zero:** No
- **Model MAE:** 28.93 (95 % CI 26.67 – 31.52)
- **Model:** 0.0793 (95 % CI 0.03715 – 0.1209)
- **Model R²:** 0.0793 (95 % CI 0.03715 – 0.1209)
- **Held-out rows:** 600
- **Held-out rows without an answer:** 0
- **Target:** Wage
- **Rows where the model is closer:** 52.67 % (95 % CI 48.67 % – 56.63 %)
- **Rows compared:** 600
- **Rows level:** 0
- **Regression yardstick trained with gaps filled (too few complete rows):** No
- **Training rows of the regression yardstick:** 2400

The model explains held-out rows about as well as training rows (held-out R² 0.0793, gap between training and held-out R² -0.0134): it **generalizes**. The structure the formula encodes carries from the training rows to the held-out rows.

- **Generalization gap:** -0.01344
- **N test:** 600
- **Target:** Wage
- **Test error:** 28.93 (95 % CI 26.67 – 31.52)
- **Test R²:** 0.0793 (95 % CI 0.03715 – 0.1209)
- **Train error:** 0.9131
- **Train R²:** 0.06586
- **Verdict:** Generalizing

### Weaker on 10 slices of the data, most in the second quarter of Logwage

*Where it is weak.* The largest miss on a held-out row is 196 (recorded 314, predicted 118). 0.3% of the held-out rows are unlike the rows it learned from.

Segment `logwage` in the second quarter (4.45 to 4.64) is **much weaker** than the aggregate (126 held-out rows): segment R² -15.2 against aggregate R² 0.0793 (gap -15.3). A drop this large means the model describes this slice of the data poorly; an input interaction or a missing column may be specific to this subset.

| Segment | Rows | R² | Gap to all rows | Verdict |
| --- | --- | --- | --- | --- |
| Logwage = q2 | 126 | -15.1960 | -15.2753 | Much weaker than on all rows |
| Logwage = q1 | 152 | -9.5522 | -9.6315 | Much weaker than on all rows |
| Logwage = q3 | 155 | -2.7989 | -2.8782 | Much weaker than on all rows |
| Education = 1_HS_GRAD | 53 | -1.7380 | -1.8173 | Much weaker than on all rows |
| Logwage = q4 | 167 | -1.0822 | -1.1615 | Much weaker than on all rows |
| Education = 5_ADVANCED_DEGREE | 86 | -0.4267 | -0.5060 | Much weaker than on all rows |
| Maritl = 4_DIVORCED | 36 | -0.2534 | -0.3327 | Much weaker than on all rows |
| Education = 2_HS_GRAD | 190 | -0.2098 | -0.2891 | Much weaker than on all rows |
| Race = 3_ASIAN | 30 | -0.1207 | -0.2000 | Weaker than on all rows |
| Age = q4 | 138 | -0.0354 | -0.1147 | Weaker than on all rows |

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

- **Aggregate R²:** 0.0793
- **Measure:** r2
- **Target:** Wage

The worst held-out row was missed by 196 (target's scale). The 10 rows with the largest absolute residuals out of 600 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.

| Actual | Predicted | Residual | Age |
| --- | --- | --- | --- |
| 314.3 | 118.0 | +196.3 | 38 |
| 309.6 | 114.0 | +195.6 | 34 |
| 276.8 | 82.0 | +194.8 | 66 |
| 311.9 | 118.0 | +193.9 | 38 |
| 268.3 | 90.0 | +178.3 | 26 |
| 276.8 | 112.0 | +164.8 | 32 |
| 277.8 | 120.0 | +157.8 | 40 |
| 276.8 | 121.0 | +155.8 | 41 |
| 281.7 | 126.0 | +155.7 | 46 |
| 272.3 | 118.0 | +154.3 | 38 |

- **Additive:** No
- **N extrapolated:** 0
- **Held-out rows:** 600
- **Held-out rows for the interval:** 600
- **Residual P5:** -50.9
- **Residual P95:** 72.97
- **Residual std:** 42.37

2 of 600 held-out rows (0.3%) lie further from the training rows than 99 in 100 rows like them would, measured on the 1 input the model reads (`age`). That is a **small minority**; the held-out figures describe the model on rows like the ones it was trained on.

| Target | Inputs measured | Line | Held-out rows | Beyond the line | Share |
| --- | --- | --- | --- | --- | --- |
| Wage | Age | 2.576 | 600 | 2 | 0.33 % |

**Score bins · 24 of 30 rows shown**

| Count | From | To |
| --- | --- | --- |
| 42 | 0 | 0.103 |
| 30 | 0.103 | 0.206 |
| 54 | 0.206 | 0.309 |
| 62 | 0.309 | 0.412 |
| 28 | 0.412 | 0.515 |
| 32 | 0.515 | 0.618 |
| 39 | 0.618 | 0.721 |
| 41 | 0.721 | 0.824 |
| 50 | 0.824 | 0.927 |
| 16 | 0.927 | 1.03 |

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

### Not ready to use yet: 1 critical concern

*What to watch.* First: Weaker quality on 10 slices of the data.

| Severity | Section | Concern | Detail |
| --- | --- | --- | --- |
| Critical | Implications | Weaker quality on 10 slices of the data | On 10 of the 32 slices compared the model is weaker than on all held-out rows (R² 0.0793). The largest gaps: Logwage in the second quarter (4.45 to 4.64) (worse than always guessing the segment's average, on 126 rows) and Logwage in the lowest quarter (3.30 to 4.44) (worse than always guessing the segment's average, on 152 rows). Predictions for these rows deserve caution; the Quality by Input Range chapter lists every slice. |
| Warning | Model | The model barely beats a straight line through the data on Wage | The model explains 7.9% of the variation on the held-out rows, where a straight line through the data explains 3.8%, a margin of 4.1 points. The formula adds little over the reference predictor; whether it is worth its complexity depends on the use. Row by row the two are about level. |
| Info | Data | Column Region holds almost one value | Region is "2_MIDDLE_ATLANTIC" in every one of its 3,000 values. A column this close to constant tells rows apart in a few cases only: a term that reads it rests on those few rows, and its charts show one bar. |
| Info | Model | Logwage reproduces Wage | Spearman's rho = 1 (Pearson r = 0.951) between Logwage 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 Logwage 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**: 3,000 rows
- **Per-Segment Predictions**: not produced, because no segment column was chosen
- **Fairness**: compared across 7 categorical columns (`education`, `health`, `health_ins`, `jobclass`, `maritl`, `race`, `region`); 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, 3 warn, 0 fail, 1 critical concern), **at least one check fails** or a critical concern is open: **Not ready**. The failing checks need addressing first; they typically indicate data leakage, severe miscalibration, or a model that loses to its baseline.

| Name | Status | Detail |
| --- | --- | --- |
| Generalization | Pass | held-out quality holds up against training quality |
| Prediction Interval Coverage | Warn | intervals are wider than their level requires, largest gap 12.2 percentage points |
| Baseline Lift for Wage | Warn | barely beats the best baseline, R² lift 0.041 |
| Model Constraints | Pass | every held-out prediction is finite and within the training range |
| Fairness | Warn | 6 columns compared, 3 flagged: Education (a group is systematically over- or under-predicted), Maritl (a group is systematically over- or under-predicted), Race (a group is systematically over- or under-predicted) |
| Out-of-Distribution | Pass | 0.3% of held-out rows are out of distribution |

- **N critical concerns:** 1
- **N fail:** 0
- **N pass:** 3
- **N warn:** 3
- **Overall verdict:** Not ready

Column `Education` splits the held-out rows into 5 groups. The prediction-parity ratio is 0.946 (smallest over largest mean prediction), the largest absolute group residual mean is 37.0 (0.873 times the overall residual standard deviation). **Residual bias**: at least one group is systematically over- or under-predicted. The model's errors are not zero-mean across groups; its predictions for the affected groups deserve caution.

- **Aggregate residual std:** 42.37
- **N columns:** 6
- **N proxies:** 0
- **Proxy threshold:** 0.5

**Columns**

| Bias max abs residual mean | Column | Groups | Interpretation | N groups | Prediction parity ratio | Verdict |
| --- | --- | --- | --- | --- | --- | --- |
| 36.97 | Education | 5_ADVANCED_DEGREE, 4_COLLEGE_GRAD, 3_SOME_COLLEGE, 2_HS_GRAD, 1_HS_GRAD | Column Education splits the held-out rows into 5 groups. The prediction-parity ratio is 0.946 (smallest over largest mean prediction), the largest absolute group residual mean is 37.0 (0.873 times the overall residual standard deviation). Residual bias: at least one group is systematically over- or under-predicted. The model's errors are not zero-mean across groups; its predictions for the affected groups deserve caution. | 5 | 0.946 | Residual bias |
| 4.702 | Health | 2_VERY_GOOD, 1_GOOD | Column Health splits the held-out rows into 2 groups. The prediction-parity ratio is 0.984 (smallest over largest mean prediction), the largest absolute group residual mean is 4.70 (0.111 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. | 2 | 0.9842 | Balanced |
| 12.58 | Health ins | 1_YES, 2_NO | Column Health ins splits the held-out rows into 2 groups. The prediction-parity ratio is 0.961 (smallest over largest mean prediction), the largest absolute group residual mean is 12.6 (0.297 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. | 2 | 0.9607 | Balanced |
| 10.31 | Jobclass | 2_INFORMATION, 1_INDUSTRIAL | Column Jobclass splits the held-out rows into 2 groups. The prediction-parity ratio is 0.978 (smallest over largest mean prediction), the largest absolute group residual mean is 10.3 (0.243 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. | 2 | 0.9782 | Balanced |
| 17.65 | Maritl | 2_MARRIED, 1_NEVER_MARRIED, 4_DIVORCED, 5_SEPARATED | Column Maritl splits the held-out rows into 4 groups. The prediction-parity ratio is 0.874 (smallest over largest mean prediction), the largest absolute group residual mean is 17.7 (0.417 times the overall residual standard deviation). Residual bias: at least one group is systematically over- or under-predicted. The model's errors are not zero-mean across groups; its predictions for the affected groups deserve caution. | 4 | 0.8737 | Residual bias |
| 22.63 | Race | 3_ASIAN, 1_WHITE, 2_BLACK, 4_OTHER | Column Race splits the held-out rows into 4 groups. The prediction-parity ratio is 0.961 (smallest over largest mean prediction), the largest absolute group residual mean is 22.6 (0.534 times the overall residual standard deviation). Residual bias: at least one group is systematically over- or under-predicted. The model's errors are not zero-mean across groups; its predictions for the affected groups deserve caution. | 4 | 0.9607 | Residual bias |

### More rows in the second quarter of Logwage would help most

*What to collect next.* There the model does worse than the average of the slice, on 126 held-out rows.

- Collect more rows where Logwage is in the lowest quarter (3.30 to 4.44), the second quarter (4.45 to 4.64), the third quarter (4.65 to 4.85), and the highest quarter (4.86 to 5.75). Reason: the model is far less accurate than overall on these segments, predicting worse than always guessing the average would for all four Logwage ranges.
- Collect more rows where Education is 1_HS_GRAD, 2_HS_GRAD, or 5_ADVANCED_DEGREE. Reason: the model is far less accurate than overall on these education groups, predicting worse than always guessing the average would.
- Collect more rows where Maritl is 4_DIVORCED. Reason: the model is far less accurate than overall on this group, predicting worse than always guessing the average would.
- Inspect the columns Education, Maritl, and Race for sources of error differences between groups. Reason: the model's errors differ between groups for these columns, as flagged by the fairness diagnostics.
- Add columns that capture more information relevant to Wage beyond Age. Reason: the model explains only 7.9% of the variation overall and barely beats a straight line through the data on Wage.

## Evidence

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

### Headline

Diafunc modeled Wage from 10 columns, using only Age, and the model explains less than half of its target: it accounts for 7.9% of Wage and relies most on Age, with which the model's output rises.

- One target was modeled; 600 rows were held out for checking.
- A straight line through the data already explained 3.8% of Wage; the model added 4.1 points, a slight and not yet settled margin.
- Quality is far weaker on 10 slices of the data, including every quarter of Logwage and several Education groups.

- **Best R²:** 0.0793
- **Best quality:** 7.38 %
- **Best target:** wage
- **Targets:** 1

### Concerns

| Severity | Section | Concern | Detail |
| --- | --- | --- | --- |
| Critical | Implications | Weaker quality on 10 slices of the data | On 10 of the 32 slices compared the model is weaker than on all held-out rows (R² 0.0793). The largest gaps: Logwage in the second quarter (4.45 to 4.64) (worse than always guessing the segment's average, on 126 rows) and Logwage in the lowest quarter (3.30 to 4.44) (worse than always guessing the segment's average, on 152 rows). Predictions for these rows deserve caution; the Quality by Input Range chapter lists every slice. |
| Warning | Model | The model barely beats a straight line through the data on Wage | The model explains 7.9% of the variation on the held-out rows, where a straight line through the data explains 3.8%, a margin of 4.1 points. The formula adds little over the reference predictor; whether it is worth its complexity depends on the use. Row by row the two are about level. |
| Info | Data | Column Region holds almost one value | Region is "2_MIDDLE_ATLANTIC" in every one of its 3,000 values. A column this close to constant tells rows apart in a few cases only: a term that reads it rests on those few rows, and its charts show one bar. |
| Info | Model | Logwage reproduces Wage | Spearman's rho = 1 (Pearson r = 0.951) between Logwage 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 Logwage 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.

### Discovered Relations

**Quality and error measured on the 600 held-out rows**

| Target | Formula | Quality | Complexity | Error |
| --- | --- | --- | --- | --- |
| Wage | Wage = bitor(80, Age) | 7.38 % | 3.5 % | 92.07 % |

*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.

### Model

The model's quality on wage is weak, just above a straight line, and all of its reliance is on age. The training quality measure is 0.0659 for share of variation explained, with held-out quality at 0.0793. There is a caution regarding model quality and logwage reproducing wage in the Concerns chapter.

Most notable in the Model section:

- `Age` 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.

#### 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.

##### Wage Model

The model's output rises with Age, which it relies on most; Age accounts for 100% of the variation explained by the formula.

```
Wage = bitor(80, Age)
```

Quality 7.38 % · Error 92.07 % on 600 held-out rows · Complexity 3.5 %

- **Held out R²:** 0.0793 (95 % CI 0.03715 – 0.1209)
- **Training rows:** 2400
- **Held-out rows:** 600
- **Inputs the formula reads:** 1
- **Terms in the formula:** 1

*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.

| Metric | Meaning |
| --- | --- |
| Quality | How 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 |
| Error | How 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 |
| Complexity | The 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.

##### 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.

###### Prediction Chart

R² of 0.0793: **weak**. 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 28.9 on the target's scale.

- **Error:** 92.07 % (95 % CI 87.91 % – 96.29 %)
- **MAE:** 28.93 (95 % CI 26.67 – 31.52)
- **N:** 600
- **Training rows:** 2400
- **Quality:** 7.38 % (95 % CI 3.14 % – 11.57 %)
- **R²:** 0.0793 (95 % CI 0.03715 – 0.1209)
- **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 colour the number of rows that fall in both, and the cells whose two ranges match hold the rows the model predicts well.

###### Residual Distribution

- **Bias:** 1.192 (95 % CI -2.278 – 4.973)
- **N:** 600
- **P50:** -4.726 (95 % CI -7.28 – -2.114)
- **P90:** 39.15 (95 % CI 34.4 – 48.7)
- **P95:** 72.97 (95 % CI 53.41 – 90.66)
- **P99:** 157.9 (95 % CI 150.3 – 194.8)
- **Std:** 42.37 (95 % CI 37.83 – 46.75)

*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.

##### Which inputs matter most

| Input | Share of importance | Permutation importance | Sensitivity | Residual correlation | Occurrences | Model's output as the input rises |
| --- | --- | --- | --- | --- | --- | --- |
| Age | 100 % | 0.1649 | 12.43 | -0.02515 | 1 | Rises (share of rows with a rising slope) |

*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.

##### Baselines

The model beats the strongest simple yardstick, a straight line through the data, by 4.1 points of variance explained on the held-out rows. Row by row the margin is **negligible**: the model reads as roughly level with the yardstick. Of the 600 held-out rows where the two differ, the model is closer on 53% (95% interval 49% to 57%). The interval of the margin includes zero: on this many held-out rows the lead is not settled.

- **Best yardstick's score:** 0.03834 (95 % CI 0.003683 – 0.06956)
- **Best baseline R²:** 0.03834
- **Effect size d:** 0.05844 (95 % CI -0.02002 – 0.1411)
- **Effect size d magnitude:** negligible
- **Margin over the best yardstick:** 0.04096 (95 % CI -0.003399 – 0.0816)
- **Margin's interval stays above zero:** No
- **Model MAE:** 28.93 (95 % CI 26.67 – 31.52)
- **Model:** 0.0793 (95 % CI 0.03715 – 0.1209)
- **Model R²:** 0.0793 (95 % CI 0.03715 – 0.1209)
- **Held-out rows:** 600
- **Held-out rows without an answer:** 0
- **Target:** Wage
- **Rows where the model is closer:** 52.67 % (95 % CI 48.67 % – 56.63 %)
- **Rows compared:** 600
- **Rows level:** 0
- **Regression yardstick trained with gaps filled (too few complete rows):** No
- **Training rows of the regression yardstick:** 2400

*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.

##### Generalization

The model explains held-out rows about as well as training rows (held-out R² 0.0793, gap between training and held-out R² -0.0134): it **generalizes**. The structure the formula encodes carries from the training rows to the held-out rows.

- **Generalization gap:** -0.01344
- **N test:** 600
- **Target:** Wage
- **Test error:** 28.93 (95 % CI 26.67 – 31.52)
- **Test R²:** 0.0793 (95 % CI 0.03715 – 0.1209)
- **Train error:** 0.9131
- **Train R²:** 0.06586
- **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.

##### How each term responds to its inputs

The formula is **not additively decomposable**: the prediction comes from a single non-separable expression. The curve below shows the whole-formula response.

- **Additive:** No
- **Terms in the formula:** 1
- **Notation:** bitor(80, Age)
- **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.

##### How the formula behaves outside the data

How the model behaves beyond the observed range of `age` could not be determined from the formula, so extrapolation there is not assessed. Extrapolate with caution.

- **Bounded at infinity:** No
- **Inputs the formula reads:** 1
- **N singularities:** 0
- **Target:** Wage

**Limits at neg infinity**

| Input | Kind |
| --- | --- |
| Age | Unknown |

**Limits at pos infinity**

| Input | Kind |
| --- | --- |
| Age | Unknown |

*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.

##### Contributions per term

- **Additive:** No
- **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.

##### What each term is worth

Full-formula R² is 0.0793. The formula is non-decomposable, so ablation collapses to a single bar; the chapter "How each term responds to its inputs" is the better surface in this case.

- **Additive:** No
- **Baseline R²:** 0.0793
- **Held-out rows:** 600

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.

##### What-If

**The prediction at the typical row: every input at its median or most common value**

| Item | Value |
| --- | --- |
| Age | 43 |
| Prediction | 123 |

- **Additive:** No
- **Held-out rows for the interval:** 600
- **Terms in the formula:** 1
- **Residual P5:** -50.9
- **Residual P95:** 72.97
- **Residual std:** 42.37

*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.

##### Sensitivity

The prediction responds to 1 input. The formula is non-additive: the change in the prediction per unit of each input (∂ŷ/∂xᵢ) is computed with the others held at the training median. **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:** No
- **Terms in the formula:** 1

*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.

##### How two inputs act together

The formula reads one input, so there is no pair of inputs that could act together. How the model's output moves with that input is in the chapter "How each term responds to its inputs".

- **Additive:** No
- **Terms in the formula:** 1

*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.

##### Worst Predictions

The worst held-out row was missed by 196 (target's scale). The 10 rows with the largest absolute residuals out of 600 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.

| Actual | Predicted | Residual | Age |
| --- | --- | --- | --- |
| 314.3 | 118.0 | +196.3 | 38 |
| 309.6 | 114.0 | +195.6 | 34 |
| 276.8 | 82.0 | +194.8 | 66 |
| 311.9 | 118.0 | +193.9 | 38 |
| 268.3 | 90.0 | +178.3 | 26 |
| 276.8 | 112.0 | +164.8 | 32 |
| 277.8 | 120.0 | +157.8 | 40 |
| 276.8 | 121.0 | +155.8 | 41 |
| 281.7 | 126.0 | +155.7 | 46 |
| 272.3 | 118.0 | +154.3 | 38 |

- **Additive:** No
- **N extrapolated:** 0
- **Held-out rows:** 600
- **Held-out rows for the interval:** 600
- **Residual P5:** -50.9
- **Residual P95:** 72.97
- **Residual std:** 42.37

*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.

##### Prediction Interval Coverage

Across 600 held-out rows, the 95% prediction interval actually covers 95.8% of the rows. The normal band is **wider than needed** for this residual distribution (worst gap 12.2 percentage points), often because a few large residuals inflate σ above what the bulk requires. Real uncertainty for typical rows is tighter than the interval suggests.

| Nominal level | Half-width | Empirical coverage |
| --- | --- | --- |
| 50 % | 28.58 | 62.17 % |
| 80 % | 54.29 | 89.83 % |
| 90 % | 69.69 | 94.5 % |
| 95 % | 83.04 | 95.83 % |

- **Largest miss (percentage points):** 12.17
- **Held-out rows:** 600
- **Residual P2.5:** -55.62
- **Residual P97.5:** 147.3
- **Verdict:** Underconfident

*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.

##### Regression Diagnostics

600 held-out residuals, with a spread up to 196 in absolute terms. Mean residual is 1.19, 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:** 0.1286
- **Bias:** 1.192
- **Bias over residual sd:** 0.02814
- **Held-out rows:** 600
- **Order:** shuffled
- **Residual sd:** 42.37
- **Share within normal band:** 0.9583

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 neighbouring 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.

##### Search Progress

The search ran for 7 generations, ending at quality 0.0732 (gain +0 from generation 3). The curve is essentially flat: either the search converged in the first few generations or the problem is fundamentally hard at this complexity limit.

- **First generation:** 3
- **First quality:** 7.32 %
- **Last generation:** 8
- **Last quality:** 7.32 %
- **Max quality:** 7.32 %
- **N points:** 7
- **N total generations:** 7

*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.

##### Model Constraints

Across 600 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.

| Name | Status | Detail |
| --- | --- | --- |
| Output coherent | Pass | no non-finite predictions across 600 rows |
| Within domain | Pass | no rows outside the training target's ±2σ range across 600 rows |
| Bounded predictions | Pass | no runaway predictions beyond ±10σ across 600 rows |

- **Held-out rows:** 600
- **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.

##### Modeling

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

- **Error:** 91.31 %
- **N train:** 2400
- **Quality:** 7.32 %

*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.

#### Findings

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

### Data

Most findings focus on how strongly education and health insurance relate to wage, along with high correlation between logwage and wage. Watch for minor issues flagged with the region column in the Concerns chapter.

Most notable in the Data section:

- `Logwage` and `Wage` move together (correlation 0.951)
- Education accounts for 23% of the spread in `Wage`: `5_ADVANCED_DEGREE` averages 66.8 above `1_HS_GRAD` (95% interval 61.1 to 72.6)
- Health ins accounts for 9.5% of the spread in `Wage`: `1_YES` averages 27.9 above `2_NO` (95% interval 25.0 to 30.8)

*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.

#### Column Profiles

Profiled 11 columns across 3,000 rows (4 numeric, 7 non-numeric): every column is at least 75% populated. Distributional statistics on numeric columns can be read at face value. 1 column is 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.

**Columns**

| Name | Role | Type | N | Missing | Mean | Std | Min | P25 | Median | P75 | Max | Skewness | CV |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Wage | Target | number | 3000 | 0 | 111.7 | 41.7 | 20.1 | 85.4 | 104.9 | 128.7 | 318.3 | 1.682 | 0.3736 |
| Age | Input | integer | 3000 | 0 | 42.41 | 11.54 | 18 | 33.75 | 42 | 51 | 80 | 0.1478 | 0.2721 |
| Education | Input | text | 3000 | 0 |  |  |  |  |  |  |  |  |  |
| Health | Input | text | 3000 | 0 |  |  |  |  |  |  |  |  |  |
| Health ins | Input | text | 3000 | 0 |  |  |  |  |  |  |  |  |  |
| Jobclass | Input | text | 3000 | 0 |  |  |  |  |  |  |  |  |  |
| Logwage | Input | number | 3000 | 0 | 4.654 | 0.352 | 3.000 | 4.447 | 4.653 | 4.857 | 5.763 | -0.1236 | 0.07558 |
| Maritl | Input | text | 3000 | 0 |  |  |  |  |  |  |  |  |  |
| Race | Input | text | 3000 | 0 |  |  |  |  |  |  |  |  |  |
| Region | Input | text | 3000 | 0 |  |  |  |  |  |  |  |  |  |

The first 10 rows of 11. The other 1 is in the notebook.

- **N columns:** 11
- **N high missing:** 0
- **N rows:** 3000
- **N strongly skewed:** 1
- **Table branch:** main
- **Table ID:** ff0ca009-3b71-4db0-9d18-21e499f5341a

*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.

#### Age

Age: 18 to 80, median 42.

*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.

##### Description

**Age**

| Statistic | Value |
| --- | --- |
| Type | integer |
| Values present | 3000 |
| Distinct values | 61 |
| Continuous | yes |
| Decimal places | 0 |
| Minimum | 18 |
| Maximum | 80 |
| Mean | 42.41 |
| Median | 42 |
| Mode | 40 |

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.

###### Correlation

**Correlations**

| Column | Pair of | Association | Pearson's r | 95% CI (Pearson) | Spearman's ρ | Kendall's τ | q-value | Rows |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Maritl | a number and a category | 0.453 |  |  |  |  | < 0.001 | 3000 |
| Logwage | two numbers | 0.2299 | 0.2179 | 0.1835 to 0.2517 | 0.2299 | 0.1583 | < 0.001 | 3000 |
| Wage | two numbers | 0.2299 | 0.1956 | 0.161 to 0.2298 | 0.2299 | 0.1583 | < 0.001 | 3000 |
| Health ins | a number and a category | 0.1426 |  |  |  |  | < 0.001 | 3000 |
| Health | a number and a category | 0.1389 |  |  |  |  | < 0.001 | 3000 |
| Education | a number and a category | 0.1074 |  |  |  |  | < 0.001 | 3000 |
| Jobclass | a number and a category | 0.09069 |  |  |  |  | < 0.001 | 3000 |
| Race | a number and a category | 0.05651 |  |  |  |  | 0.02526 | 3000 |
| Year | two numbers | 0.03867 | 0.03843 | 0.002641 to 0.07411 | 0.03867 | 0.02814 | 0.03533 | 3000 |

*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.

| Measure | Meaning |
| --- | --- |
| Kind | What 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 |
| Association | The 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 |
| Covariance | Numeric 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 Coefficient | Numeric 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 ρ | Numeric 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 τ | Numeric 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-value | How 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 |
| Rows | The rows in which both columns hold a value. The q-value and the interval of a pair rest on this count |

##### 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 0.01 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.

#### Education

Education: 5 distinct values, most often `2_HS_GRAD`.

##### Description

**Education**

| Statistic | Value |
| --- | --- |
| Type | text |
| Values present | 3000 |
| Distinct values | 5 |
| Continuous | no |
| Mode | 2_HS_GRAD |

**Rows per value**

| Variable | Value | Rows | Share |
| --- | --- | --- | --- |
| Education | 2_HS_GRAD | 971 | 32.37 % |
| Education | 4_COLLEGE_GRAD | 685 | 22.83 % |
| Education | 3_SOME_COLLEGE | 650 | 21.67 % |
| Education | 5_ADVANCED_DEGREE | 426 | 14.2 % |
| Education | 1_HS_GRAD | 268 | 8.933 % |

###### By Value

Education accounts for 23% of the spread in `Wage`: `5_ADVANCED_DEGREE` averages 66.8 above `1_HS_GRAD` (95% interval 61.1 to 72.6). An association, not a cause.

**Groups**

| Target | Value | Rows | Average or share | 95% interval |
| --- | --- | --- | --- | --- |
| Wage | 5_ADVANCED_DEGREE | 426 | 150.9 | 145.8 to 156.1 |
| Wage | 4_COLLEGE_GRAD | 685 | 124.4 | 121.3 to 127.5 |
| Wage | 3_SOME_COLLEGE | 650 | 107.8 | 105.3 to 110.3 |
| Wage | 2_HS_GRAD | 971 | 95.78 | 93.98 to 97.58 |
| Wage | 1_HS_GRAD | 268 | 84.1 | 81.51 to 86.7 |

**Mean per value of Education**

| Education | Rows | Logwage | Year | Age | Wage |
| --- | --- | --- | --- | --- | --- |
| 1_HS_GRAD | 268 | 4.398 | 2005.8 | 41.79 | 84.1 |
| 2_HS_GRAD | 971 | 4.521 | 2005.8 | 42.22 | 95.78 |
| 3_SOME_COLLEGE | 650 | 4.636 | 2005.7 | 40.89 | 107.8 |
| 4_COLLEGE_GRAD | 685 | 4.771 | 2005.8 | 42.77 | 124.4 |
| 5_ADVANCED_DEGREE | 426 | 4.958 | 2005.9 | 45.01 | 150.9 |

*How to read this*

> The rows are grouped by the value of `Education`, 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 `Education`; 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.

#### Health

Health: 2 distinct values, most often `2_VERY_GOOD`.

##### Description

**Health**

| Statistic | Value |
| --- | --- |
| Type | text |
| Values present | 3000 |
| Distinct values | 2 |
| Continuous | no |
| Mode | 2_VERY_GOOD |

**Rows per value**

| Variable | Value | Rows | Share |
| --- | --- | --- | --- |
| Health | 2_VERY_GOOD | 2142 | 71.4 % |
| Health | 1_GOOD | 858 | 28.6 % |

###### By Value

**Mean per value of Health**

| Health | Rows | Logwage | Year | Age | Wage |
| --- | --- | --- | --- | --- | --- |
| 1_GOOD | 858 | 4.566 | 2005.8 | 44.95 | 101.7 |
| 2_VERY_GOOD | 2142 | 4.689 | 2005.8 | 41.4 | 115.7 |

#### Health ins

Health ins: 2 distinct values, most often `1_YES`.

##### Description

**Health ins**

| Statistic | Value |
| --- | --- |
| Type | text |
| Values present | 3000 |
| Distinct values | 2 |
| Continuous | no |
| Mode | 1_YES |

**Rows per value**

| Variable | Value | Rows | Share |
| --- | --- | --- | --- |
| Health ins | 1_YES | 2083 | 69.43 % |
| Health ins | 2_NO | 917 | 30.57 % |

###### By Value

Health ins accounts for 9.5% of the spread in `Wage`: `1_YES` averages 27.9 above `2_NO` (95% interval 25.0 to 30.8). An association, not a cause.

**Groups**

| Target | Value | Rows | Average or share | 95% interval |
| --- | --- | --- | --- | --- |
| Wage | 1_YES | 2083 | 120.2 | 118.5 to 122 |
| Wage | 2_NO | 917 | 92.32 | 89.99 to 94.65 |

**Mean per value of Health ins**

| Health ins | Rows | Logwage | Year | Age | Wage |
| --- | --- | --- | --- | --- | --- |
| 1_YES | 2083 | 4.74 | 2005.8 | 43.51 | 120.2 |
| 2_NO | 917 | 4.458 | 2005.8 | 39.93 | 92.32 |

#### Jobclass

Jobclass: 2 distinct values, most often `1_INDUSTRIAL`.

##### Description

**Jobclass**

| Statistic | Value |
| --- | --- |
| Type | text |
| Values present | 3000 |
| Distinct values | 2 |
| Continuous | no |
| Mode | 1_INDUSTRIAL |

**Rows per value**

| Variable | Value | Rows | Share |
| --- | --- | --- | --- |
| Jobclass | 1_INDUSTRIAL | 1544 | 51.47 % |
| Jobclass | 2_INFORMATION | 1456 | 48.53 % |

###### By Value

**Mean per value of Jobclass**

| Jobclass | Rows | Logwage | Year | Age | Wage |
| --- | --- | --- | --- | --- | --- |
| 1_INDUSTRIAL | 1544 | 4.584 | 2005.8 | 41.4 | 103.3 |
| 2_INFORMATION | 1456 | 4.728 | 2005.8 | 43.49 | 120.6 |

#### Logwage

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.

Logwage: 3 to 5.76, median 4.65; moves with `Wage` most (r 0.951).

##### Description

**Logwage**

| Statistic | Value |
| --- | --- |
| Type | number |
| Values present | 3000 |
| Distinct values | 508 |
| Continuous | yes |
| Decimal places | 14 |
| Minimum | 3 |
| Maximum | 5.763 |
| Mean | 4.654 |
| Median | 4.653 |
| Mode | 4.77815125038364 |

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

###### Correlation

**Correlations**

| Column | Pair of | Association | Pearson's r | 95% CI (Pearson) | Spearman's ρ | Kendall's τ | q-value | Rows |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Wage | two numbers | 1 | 0.9507 | 0.9471 to 0.954 | 1 | 1 | < 0.001 | 3000 |
| Education | a number and a category | 0.4756 |  |  |  |  | < 0.001 | 3000 |
| Health ins | a number and a category | 0.3697 |  |  |  |  | < 0.001 | 3000 |
| Maritl | a number and a category | 0.2898 |  |  |  |  | < 0.001 | 3000 |
| Age | two numbers | 0.2299 | 0.2179 | 0.1835 to 0.2517 | 0.2299 | 0.1583 | < 0.001 | 3000 |
| Jobclass | a number and a category | 0.2054 |  |  |  |  | < 0.001 | 3000 |
| Health | a number and a category | 0.1583 |  |  |  |  | < 0.001 | 3000 |
| Race | a number and a category | 0.1106 |  |  |  |  | < 0.001 | 3000 |
| Year | two numbers | 0.07624 | 0.07624 | 0.04056 to 0.1117 | 0.07518 | 0.05433 | < 0.001 | 3000 |

##### Visualization

The rows of `Logwage` 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.

#### Maritl

Maritl: 5 distinct values, most often `2_MARRIED`.

##### Description

**Maritl**

| Statistic | Value |
| --- | --- |
| Type | text |
| Values present | 3000 |
| Distinct values | 5 |
| Continuous | no |
| Mode | 2_MARRIED |

**Rows per value**

| Variable | Value | Rows | Share |
| --- | --- | --- | --- |
| Maritl | 2_MARRIED | 2074 | 69.13 % |
| Maritl | 1_NEVER_MARRIED | 648 | 21.6 % |
| Maritl | 4_DIVORCED | 204 | 6.8 % |
| Maritl | 5_SEPARATED | 55 | 1.833 % |
| Maritl | 3_WIDOWED | 19 | 0.6333 % |

###### By Value

Maritl accounts for 6.8% of the spread in `Wage`: `2_MARRIED` averages 26.1 above `1_NEVER_MARRIED` (95% interval 23.0 to 29.3). An association, not a cause.

**Groups**

| Target | Value | Rows | Average or share | 95% interval |
| --- | --- | --- | --- | --- |
| Wage | 2_MARRIED | 2074 | 118.9 | 117 to 120.7 |
| Wage | 4_DIVORCED | 204 | 103.2 | 98.49 to 107.8 |
| Wage | 5_SEPARATED | 55 | 101.2 | 92.12 to 110.3 |
| Wage | 3_WIDOWED | 19 | 99.54 | 88.1 to 111 |
| Wage | 1_NEVER_MARRIED | 648 | 92.73 | 90.2 to 95.27 |

**Mean per value of Maritl**

| Maritl | Rows | Logwage | Year | Age | Wage |
| --- | --- | --- | --- | --- | --- |
| 1_NEVER_MARRIED | 648 | 4.475 | 2005.9 | 32.58 | 92.73 |
| 2_MARRIED | 2074 | 4.72 | 2005.8 | 44.86 | 118.9 |
| 3_WIDOWED | 19 | 4.571 | 2005.2 | 50.79 | 99.54 |
| 4_DIVORCED | 204 | 4.584 | 2005.7 | 47.61 | 103.2 |
| 5_SEPARATED | 55 | 4.577 | 2006 | 43.82 | 101.2 |

#### Race

Race: 4 distinct values, most often `1_WHITE`.

##### Description

**Race**

| Statistic | Value |
| --- | --- |
| Type | text |
| Values present | 3000 |
| Distinct values | 4 |
| Continuous | no |
| Mode | 1_WHITE |

**Rows per value**

| Variable | Value | Rows | Share |
| --- | --- | --- | --- |
| Race | 1_WHITE | 2480 | 82.67 % |
| Race | 2_BLACK | 293 | 9.767 % |
| Race | 3_ASIAN | 190 | 6.333 % |
| Race | 4_OTHER | 37 | 1.233 % |

###### By Value

**Mean per value of Race**

| Race | Rows | Logwage | Year | Age | Wage |
| --- | --- | --- | --- | --- | --- |
| 1_WHITE | 2480 | 4.662 | 2005.8 | 42.39 | 112.6 |
| 2_BLACK | 293 | 4.571 | 2005.9 | 43.56 | 101.6 |
| 3_ASIAN | 190 | 4.72 | 2006 | 41.82 | 120.3 |
| 4_OTHER | 37 | 4.451 | 2005.8 | 37.7 | 89.97 |

#### Region

Region: 1 distinct value, most often `2_MIDDLE_ATLANTIC`.

##### Description

**Region**

| Statistic | Value |
| --- | --- |
| Type | text |
| Values present | 3000 |
| Distinct values | 1 |
| Continuous | no |
| Mode | 2_MIDDLE_ATLANTIC |

**Rows per value**

| Variable | Value | Rows | Share |
| --- | --- | --- | --- |
| Region | 2_MIDDLE_ATLANTIC | 3000 | 100 % |

#### Wage

Wage: 20.1 to 318, median 105; moves with `Logwage` most (r 0.951).

##### Description

**Wage**

| Statistic | Value |
| --- | --- |
| Type | number |
| Values present | 3000 |
| Distinct values | 508 |
| Continuous | yes |
| Decimal places | 15 |
| Minimum | 20.09 |
| Maximum | 318.3 |
| Mean | 111.7 |
| Median | 104.9 |
| Mode | 118.88435933988579 |

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

###### Correlation

**Correlations**

| Column | Pair of | Association | Pearson's r | 95% CI (Pearson) | Spearman's ρ | Kendall's τ | q-value | Rows |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Logwage | two numbers | 1 | 0.9507 | 0.9471 to 0.954 | 1 | 1 | < 0.001 | 3000 |
| Education | a number and a category | 0.4846 |  |  |  |  | < 0.001 | 3000 |
| Health ins | a number and a category | 0.3083 |  |  |  |  | < 0.001 | 3000 |
| Maritl | a number and a category | 0.2637 |  |  |  |  | < 0.001 | 3000 |
| Age | two numbers | 0.2299 | 0.1956 | 0.161 to 0.2298 | 0.2299 | 0.1583 | < 0.001 | 3000 |
| Jobclass | a number and a category | 0.2069 |  |  |  |  | < 0.001 | 3000 |
| Health | a number and a category | 0.1523 |  |  |  |  | < 0.001 | 3000 |
| Race | a number and a category | 0.11 |  |  |  |  | < 0.001 | 3000 |
| Year | two numbers | 0.07518 | 0.06554 | 0.02983 to 0.1011 | 0.07518 | 0.05433 | < 0.001 | 3000 |

##### 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 -0.01 between each value and the next), so the order carries no pattern to show.

#### Year

Year: 2,003 to 2,009, median 2,006, 7 distinct values.

##### Description

**Year**

| Statistic | Value |
| --- | --- |
| Type | integer |
| Values present | 3000 |
| Distinct values | 7 |
| Continuous | no |
| Decimal places | 0 |
| Minimum | 2003 |
| Maximum | 2009 |
| Mean | 2005.8 |
| Median | 2006 |
| Mode | 2003 |

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

**Rows per value**

| Variable | Value | Rows | Share |
| --- | --- | --- | --- |
| Year | 2003 | 513 | 17.1 % |
| Year | 2004 | 485 | 16.17 % |
| Year | 2005 | 447 | 14.9 % |
| Year | 2006 | 392 | 13.07 % |
| Year | 2007 | 386 | 12.87 % |
| Year | 2008 | 388 | 12.93 % |
| Year | 2009 | 389 | 12.97 % |

###### By Value

**Mean per value of Year**

| Year | Rows | Logwage | Age | Wage |
| --- | --- | --- | --- | --- |
| 2003 | 513 | 4.594 | 42.06 | 106.2 |
| 2004 | 485 | 4.649 | 42.33 | 111.2 |
| 2005 | 447 | 4.645 | 41.58 | 110 |
| 2006 | 392 | 4.68 | 42.02 | 114.2 |
| 2007 | 386 | 4.671 | 43.04 | 112.9 |
| 2008 | 388 | 4.669 | 42.54 | 113.6 |
| 2009 | 389 | 4.691 | 43.59 | 116 |

###### Correlation

**Correlations**

| Column | Pair of | Association | Pearson's r | 95% CI (Pearson) | Spearman's ρ | Kendall's τ | q-value | Rows |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Logwage | two numbers | 0.07624 | 0.07624 | 0.04056 to 0.1117 | 0.07518 | 0.05433 | < 0.001 | 3000 |
| Wage | two numbers | 0.07518 | 0.06554 | 0.02983 to 0.1011 | 0.07518 | 0.05433 | 0.001475 | 3000 |
| Race | a number and a category | 0.03887 |  |  |  |  | 0.4716 | 3000 |
| Age | two numbers | 0.03867 | 0.03843 | 0.002641 to 0.07411 | 0.03867 | 0.02814 | 0.106 | 3000 |
| Maritl | a number and a category | 0.03309 |  |  |  |  | 0.8269 | 3000 |
| Education | a number and a category | 0.03184 |  |  |  |  | 0.8269 | 3000 |
| Health ins | a number and a category | 0.008091 |  |  |  |  | 0.8281 | 3000 |
| Jobclass | a number and a category | 0.006155 |  |  |  |  | 0.8281 | 3000 |
| Health | a number and a category | 0.001938 |  |  |  |  | 0.9155 | 3000 |

##### Visualization

The rows of `Year` 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.

#### Column · Region

`Region` is profiled on its own because it may hold free text.

Scanned 3,000 non-null values in `region`, extracted 6,000 content tokens (2 unique) after dropping 0 English stopword instances. The most frequent content word is `atlantic` (3,000 occurrences, 50% of all kept tokens). The table maps the column's content vocabulary; a heavy concentration on one or two tokens may signal a near-constant column or a categorical disguised as free text. The stopword filter is **English-only**: in a non-English column its own function words dominate the top.

_This section continues in the notebook._

### Implications

The weakest accuracy is found on rows with logwage in the lower to middle range, suggesting the formula is poorly suited for those slices. Some critical quality issues are flagged in the Concerns chapter.

Most notable in the Implications section:

- The model is far less accurate on rows with `Logwage` in the second quarter (4.45 to 4.64) (worse than always predicting the average of those rows)
- The model is far less accurate on rows with `Logwage` in the lowest quarter (3.30 to 4.44) (worse than always predicting the average of those rows)
- The model is far less accurate on rows with `Logwage` in the third quarter (4.65 to 4.85) (worse than always predicting the average of those rows)

*How to read this*

> 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.

#### Stakeholder Summary

The model answers how Wage moves with Age in this data, describing only a small part of the variation, and only just ahead of a straight-line trend. It tracks Age alone, when Age rises, the model's output rises as well, but this is a slim advance, not enough to show consistent gains row by row. The margin over a straight line is slight and not yet settled as a clear lead.

To rely on the model’s output for individual predictions or decisions is not advised. Its answer captures only a minor pattern, and it is far less accurate for every quarter of Logwage and several Education groups. Results differ between groups for Education, Maritl, and Race. Validate on a fresh sample before using the model for decisions.

#### Recommended Next Data

- Collect more rows where Logwage is in the lowest quarter (3.30 to 4.44), the second quarter (4.45 to 4.64), the third quarter (4.65 to 4.85), and the highest quarter (4.86 to 5.75). Reason: the model is far less accurate than overall on these segments, predicting worse than always guessing the average would for all four Logwage ranges.
- Collect more rows where Education is 1_HS_GRAD, 2_HS_GRAD, or 5_ADVANCED_DEGREE. Reason: the model is far less accurate than overall on these education groups, predicting worse than always guessing the average would.
- Collect more rows where Maritl is 4_DIVORCED. Reason: the model is far less accurate than overall on this group, predicting worse than always guessing the average would.
- Inspect the columns Education, Maritl, and Race for sources of error differences between groups. Reason: the model's errors differ between groups for these columns, as flagged by the fairness diagnostics.
- Add columns that capture more information relevant to Wage beyond Age. Reason: the model explains only 7.9% of the variation overall and barely beats a straight line through the data on Wage.

#### Quality by Input Range

Quality by Input Range splits the held-out rows by every usable input and compares the model's quality on each segment with its quality on all held-out rows. 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 input is split into quartiles; a categorical input 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.

##### Wage by Input Range

Segment `logwage` in the second quarter (4.45 to 4.64) is **much weaker** than the aggregate (126 held-out rows): segment R² -15.2 against aggregate R² 0.0793 (gap -15.3). A drop this large means the model describes this slice of the data poorly; an input interaction or a missing column may be specific to this subset.

| Segment | Rows | R² | Gap to all rows | Verdict |
| --- | --- | --- | --- | --- |
| Logwage = q2 | 126 | -15.1960 | -15.2753 | Much weaker than on all rows |
| Logwage = q1 | 152 | -9.5522 | -9.6315 | Much weaker than on all rows |
| Logwage = q3 | 155 | -2.7989 | -2.8782 | Much weaker than on all rows |
| Education = 1_HS_GRAD | 53 | -1.7380 | -1.8173 | Much weaker than on all rows |
| Logwage = q4 | 167 | -1.0822 | -1.1615 | Much weaker than on all rows |
| Education = 5_ADVANCED_DEGREE | 86 | -0.4267 | -0.5060 | Much weaker than on all rows |
| Maritl = 4_DIVORCED | 36 | -0.2534 | -0.3327 | Much weaker than on all rows |
| Education = 2_HS_GRAD | 190 | -0.2098 | -0.2891 | Much weaker than on all rows |
| Race = 3_ASIAN | 30 | -0.1207 | -0.2000 | Weaker than on all rows |
| Age = q4 | 138 | -0.0354 | -0.1147 | Weaker than on all rows |

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

- **Aggregate R²:** 0.0793
- **Measure:** r2
- **Target:** Wage

#### Rows unlike the ones the model learned from

2 of 600 held-out rows (0.3%) lie further from the training rows than 99 in 100 rows like them would, measured on the 1 input the model reads (`age`). That is a **small minority**; the held-out figures describe the model on rows like the ones it was trained on.

| Target | Inputs measured | Line | Held-out rows | Beyond the line | Share |
| --- | --- | --- | --- | --- | --- |
| Wage | Age | 2.576 | 600 | 2 | 0.33 % |

**Score bins · 24 of 30 rows shown**

| Count | From | To |
| --- | --- | --- |
| 42 | 0 | 0.103 |
| 30 | 0.103 | 0.206 |
| 54 | 0.206 | 0.309 |
| 62 | 0.309 | 0.412 |
| 28 | 0.412 | 0.515 |
| 32 | 0.515 | 0.618 |
| 39 | 0.618 | 0.721 |
| 41 | 0.721 | 0.824 |
| 50 | 0.824 | 0.927 |
| 16 | 0.927 | 1.03 |

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.

#### 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.

##### Wage Fairness

Column `Education` splits the held-out rows into 5 groups. The prediction-parity ratio is 0.946 (smallest over largest mean prediction), the largest absolute group residual mean is 37.0 (0.873 times the overall residual standard deviation). **Residual bias**: at least one group is systematically over- or under-predicted. The model's errors are not zero-mean across groups; its predictions for the affected groups deserve caution.

- **Aggregate residual std:** 42.37
- **N columns:** 6
- **N proxies:** 0
- **Proxy threshold:** 0.5

**Columns**

| Bias max abs residual mean | Column | Groups | Interpretation | N groups | Prediction parity ratio | Verdict |
| --- | --- | --- | --- | --- | --- | --- |
| 36.97 | Education | 5_ADVANCED_DEGREE, 4_COLLEGE_GRAD, 3_SOME_COLLEGE, 2_HS_GRAD, 1_HS_GRAD | Column Education splits the held-out rows into 5 groups. The prediction-parity ratio is 0.946 (smallest over largest mean prediction), the largest absolute group residual mean is 37.0 (0.873 times the overall residual standard deviation). Residual bias: at least one group is systematically over- or under-predicted. The model's errors are not zero-mean across groups; its predictions for the affected groups deserve caution. | 5 | 0.946 | Residual bias |
| 4.702 | Health | 2_VERY_GOOD, 1_GOOD | Column Health splits the held-out rows into 2 groups. The prediction-parity ratio is 0.984 (smallest over largest mean prediction), the largest absolute group residual mean is 4.70 (0.111 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. | 2 | 0.9842 | Balanced |
| 12.58 | Health ins | 1_YES, 2_NO | Column Health ins splits the held-out rows into 2 groups. The prediction-parity ratio is 0.961 (smallest over largest mean prediction), the largest absolute group residual mean is 12.6 (0.297 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. | 2 | 0.9607 | Balanced |
| 10.31 | Jobclass | 2_INFORMATION, 1_INDUSTRIAL | Column Jobclass splits the held-out rows into 2 groups. The prediction-parity ratio is 0.978 (smallest over largest mean prediction), the largest absolute group residual mean is 10.3 (0.243 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. | 2 | 0.9782 | Balanced |
| 17.65 | Maritl | 2_MARRIED, 1_NEVER_MARRIED, 4_DIVORCED, 5_SEPARATED | Column Maritl splits the held-out rows into 4 groups. The prediction-parity ratio is 0.874 (smallest over largest mean prediction), the largest absolute group residual mean is 17.7 (0.417 times the overall residual standard deviation). Residual bias: at least one group is systematically over- or under-predicted. The model's errors are not zero-mean across groups; its predictions for the affected groups deserve caution. | 4 | 0.8737 | Residual bias |
| 22.63 | Race | 3_ASIAN, 1_WHITE, 2_BLACK, 4_OTHER | Column Race splits the held-out rows into 4 groups. The prediction-parity ratio is 0.961 (smallest over largest mean prediction), the largest absolute group residual mean is 22.6 (0.534 times the overall residual standard deviation). Residual bias: at least one group is systematically over- or under-predicted. The model's errors are not zero-mean across groups; its predictions for the affected groups deserve caution. | 4 | 0.9607 | Residual bias |

#### 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.

**Upgrade pathways**

| Name | Summary |
| --- | --- |
| Controlled experiment | Randomize the input to be acted on, then analyze again. Randomization breaks confounding, so any change in the target is attributable to the input. |
| Natural experiment | Find a setting where the input varied for reasons unrelated to the target, an instrumental variable. Variation of that kind supports causal inference. |
| Causal graph | Write out what plausibly causes what, then check whether the model's terms are consistent with the conditional independences the graph implies. |
| Control for confounders | If 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. |

#### 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**: 3,000 rows
- **Per-Segment Predictions**: not produced, because no segment column was chosen
- **Fairness**: compared across 7 categorical columns (`education`, `health`, `health_ins`, `jobclass`, `maritl`, `race`, `region`); 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, 3 warn, 0 fail, 1 critical concern), **at least one check fails** or a critical concern is open: **Not ready**. The failing checks need addressing first; they typically indicate data leakage, severe miscalibration, or a model that loses to its baseline.

| Name | Status | Detail |
| --- | --- | --- |
| Generalization | Pass | held-out quality holds up against training quality |
| Prediction Interval Coverage | Warn | intervals are wider than their level requires, largest gap 12.2 percentage points |
| Baseline Lift for Wage | Warn | barely beats the best baseline, R² lift 0.041 |
| Model Constraints | Pass | every held-out prediction is finite and within the training range |
| Fairness | Warn | 6 columns compared, 3 flagged: Education (a group is systematically over- or under-predicted), Maritl (a group is systematically over- or under-predicted), Race (a group is systematically over- or under-predicted) |
| Out-of-Distribution | Pass | 0.3% of held-out rows are out of distribution |

- **N critical concerns:** 1
- **N fail:** 0
- **N pass:** 3
- **N warn:** 3
- **Overall verdict:** Not ready

#### 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.

| Item | Value |
| --- | --- |
| Analysis ID | 0a807890-48b8-11f1-8087-d378640fe787 |
| Run time | 2026-10-24 17:42 UTC |
| Total rows | 3,000 |
| Modeled targets | Wage |
| Segment column | not available |
| Time limit | 1 h 0 min |
| Time taken | 1 min 33 s |
| Credit budget | 500 credits |
| Models kept from | run 944af700-b6b6-11f1-a831-412b3a9c9ef7, whose search found them; this report was built again around them |
| Left out of the inputs | Logwage |

#### Findings

1 finding from the implications diagnostics (1 Critical) is listed in the Concerns chapter at the top of the analysis.

_The report continues in the notebook._
