PDP report
target charges · 6 features · 3 plotted
1 · Overview — where to look
An additive surrogate read off the global curves reproduces 84.4% of the model's predicted variance; adding the regional plots kept by the decision sequence, 96.6%.
Each point is a feature: importance (x) against heterogeneity (y). Bottom-left is ignorable; bottom-right is important and fully described by its mean effect; the top-right corner — important and heterogeneous — is where the mean hides something. An arrow marks each split the decision sequence accepted: from the feature's global point to its weighted-mean point across the subregions.
| # | feature | importance | heterogeneity | #regions | regional analysis |
|---|---|---|---|---|---|
| 1 | smoker | 9741.7124 | 4202.7570 | 7 | split found — rejected by the decision sequence → |
| 2 | age | 3665.6089 | 1547.5959 | 7 | split found — rejected by the decision sequence → |
| 3 | bmi | 2669.0218 | 1229.4831 | 7 | split into 4 regions → |
| 4 | children | 939.8014 | 1057.1503 | · | not plotted (below the coverage cut) |
| 5 | region | 481.7119 | 903.0898 | · | not plotted (below the coverage cut) |
| 6 | sex | 198.2283 | 483.7861 | · | not plotted (below the coverage cut) |
The plotted features carry 91% of the total importance mass (target 80%, ceiling top_k = 5).
The decision sequence. Starting from the global curves, each round applies the split with the largest explained-variance gain, measured on top of the splits above it, and stops when no remaining split adds at least 1.0%. A real split (its heterogeneity does drop) can still add nothing — or even hurt, by double-counting — when its variance is already explained by an earlier split.
| step | regions | heterogeneity | explained variance |
|---|---|---|---|
| global effects (GAM) | · | · | 84.4% |
| + split bmi (on age, smoker) | 4 | 4497.234 → 1229.483 | +12.1% → 96.6% |
| rejected · smoker (on bmi) | 4 | 4202.757 → 1421.014 | -11.5% — redundant (variance already explained) |
| rejected · age (on bmi, children, smoker) | 4 | 1547.596 → 1312.774 | +0.4% — below the 1.0% threshold |
Bar view — importance and heterogeneity
2 · Regional analysis — the final CALM
The selected snapshot: global effects everywhere except the accepted splits. Features in descending importance — a split feature enters as one group at the instance-weighted mean of its subregions. The split features' global counterparts are in the baseline section at the end.
2.1 · smoker
Global effect
Regional effects
A split on bmi into 4 regions was found (heterogeneity 4202.757 → 1421.014), but the decision sequence skips it: it adds no explained variance beyond the splits kept there — the same variance is already read elsewhere. The regional plots are omitted; reproduce them with find_regions.
2.2 · age
Global effect
Regional effects
A split on bmi, children, smoker into 4 regions was found (heterogeneity 1547.596 → 1312.774), but the decision sequence skips it: it adds only +0.4%, below the 1.0% threshold. The regional plots are omitted; reproduce them with find_regions.
2.3 · bmi
Split on age, smoker into 4 regions — worth +12.1% of explained variance on top of the splits above it; importance and heterogeneity here are the instance-weighted means over the subregions. The global counterpart is in the baseline.
Partition tree
Feature 2 - Full partition tree:
🌳 Full Tree Structure:
───────────────────────
bmi 🔹 [id: 0 | heter: 4497.23 | inst: 1070 | w: 1.00]
smoker = no 🔹 [id: 1 | heter: 1439.97 | inst: 841 | w: 0.79]
age < 43.30 🔹 [id: 2 | heter: 1162.94 | inst: 485 | w: 0.45]
age ≥ 43.30 🔹 [id: 3 | heter: 1423.27 | inst: 356 | w: 0.33]
smoker = yes 🔹 [id: 4 | heter: 1224.35 | inst: 229 | w: 0.21]
age < 41.00 🔹 [id: 5 | heter: 1111.57 | inst: 125 | w: 0.12]
age ≥ 41.00 🔹 [id: 6 | heter: 1018.18 | inst: 104 | w: 0.10]
--------------------------------------------------
Feature 2 - Statistics per tree level:
🌳 Tree Summary:
─────────────────
Level 0🔹heter: 4497.23
Level 1🔹heter: 1393.82 | 🔻3103.41 (69.01%)
Level 2🔹heter: 1229.48 | 🔻164.34 (11.79%)
3 · Global baseline — without regions
What you would believe about the split features without the regional analysis: their global mean effects, with the heterogeneity the accepted splits just explained still hiding inside the band. Compare with their subregions in the regional analysis above.
bmi
