PDP report
target y · 3 features · 2 plotted
1 · Overview — where to look
An additive surrogate read off the global curves reproduces 2.7% of the model's predicted variance; adding the regional plots kept by the decision sequence, 100.0%.
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 | x0 | 0.8794 | 0.0000 | 7 | split into 4 regions → |
| 2 | x1 | 0.2141 | 1.2365 | 5 | split found — rejected by the decision sequence → |
| 3 | x2 | 0.0236 | 0.8672 | · | not plotted (below the coverage cut) |
The plotted features carry 98% of the total importance mass (target 80%, ceiling top_k = 3).
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) | · | · | 2.7% |
| + split x0 (on x1, x2) | 4 | 1.236 → 0.000 | +97.3% → 100.0% |
| rejected · x1 (on x0, x2) | 3 | 1.237 → 0.441 | -71.0% — redundant (variance already explained) |
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 · x0
Split on x1, x2 into 4 regions — worth +97.3% 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 0 - Full partition tree:
🌳 Full Tree Structure:
───────────────────────
x0 🔹 [id: 0 | heter: 1.24 | inst: 3000 | w: 1.00]
x1 < 0.00 🔹 [id: 1 | heter: 0.89 | inst: 1480 | w: 0.49]
x2 = 0.00 🔹 [id: 2 | heter: 0.00 | inst: 742 | w: 0.25]
x2 = 1.00 🔹 [id: 3 | heter: 0.00 | inst: 738 | w: 0.25]
x1 ≥ 0.00 🔹 [id: 4 | heter: 0.85 | inst: 1520 | w: 0.51]
x2 = 0.00 🔹 [id: 5 | heter: 0.00 | inst: 778 | w: 0.26]
x2 = 1.00 🔹 [id: 6 | heter: 0.00 | inst: 742 | w: 0.25]
--------------------------------------------------
Feature 0 - Statistics per tree level:
🌳 Tree Summary:
─────────────────
Level 0🔹heter: 1.24
Level 1🔹heter: 0.87 | 🔻0.37 (29.77%)
Level 2🔹heter: 0.00 | 🔻0.87 (100.00%)
2.2 · x1
Global effect
Regional effects
A split on x0, x2 into 3 regions was found (heterogeneity 1.237 → 0.441), 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.
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.
x0
