Follow the transformation
Start with the audience and decision.
Choose an explanation method that matches model type and question (global vs local).
Use held-out data when measuring importance that depends on predictive performance.
Interpretation & Communication · Lesson 76
PDP and ICE
PDP and ICE is part of model interpretation and communication.
What PDP and ICE actually means
PDP and ICE is part of model interpretation and communication. Explanations describe model behaviour under specific assumptions; they are not automatically causal statements about the real world.
PDP and ICE matters because interpretation is useful only after predictive validity is established and only within the assumptions of the explanation method. Communication must distinguish model reliance, statistical association and causal effect.
Deeper walkthrough
Read PDP and ICE as a mechanism, not a recipe
Treat this as a sequence of observable decisions rather than one opaque command. Stage 1: Start with the audience and decision. Stage 2: Choose an explanation method that matches model type and question (global vs local). Stage 3: Use held-out data when measuring importance that depends on predictive performance. Final checkpoint: State uncertainty, limitations and what the explanation does not establish.
Know what would convince you
- Repeat the explanation on held-out/subsampled data and check whether the important pattern is stable.
- Compare at least one alternative explanation or direct prediction perturbation for a small case.
Useful distinctionCoefficients: Direct model parameters; interpretation depends on scaling/encoding and model form.
Click a stage to inspect what happens, what changes, and what should be checked before moving on.
Stage 1
Start with the audience and decision
Start with the audience and decision. For PDP and ICE, identify the exact state before this stage, the operation or rule applied here, and the observable state afterwards so the mechanism remains inspectable.
State focus: identify exactly what changed at this stage and what observable evidence confirms that change.
Trace the mechanism step by step
- Start with the audience and decision.
- Choose an explanation method that matches model type and question (global vs local).
- Use held-out data when measuring importance that depends on predictive performance.
- Check correlated features because importance can be shared or displaced.
- State uncertainty, limitations and what the explanation does not establish.
Make the concept concrete
Demonstration
Text example
PDP: average the model prediction after setting a feature to each grid value across all rows.
ICE: draw the same prediction-vs-feature curve separately for individual rows.Expected / illustrative result
If ICE curves differ strongly, a single average PDP can hide heterogeneous effects or interactions.
Interpret the result.
For PDP and ICE, connect the displayed result to the specific input and mechanism above; independently verify one value/state change rather than treating successful execution as proof.
Know what this is — and what it is not
CoefficientsDirect model parameters; interpretation depends on scaling/encoding and model form.Permutation importancePerformance loss after shuffling a feature.PDPAverage predicted response as a feature is varied, marginalising over data.ICEIndividual prediction curves rather than the average.SHAPAdditive feature-attribution framework tied to a background/reference distribution.When it is appropriate
Use PDP and ICE when the explanation question is explicit—global behaviour, local prediction, feature effect or communication—and the method’s limitations are acceptable.
When to stop or reconsider
Do not treat model explanations as causal effects or ground truth, especially with correlated features, extrapolation or unstable models.
Failure modes to recognise
- Presenting feature importance or attribution as causality.
- Ignoring correlated/interacting features that can redistribute or mask apparent importance.
- Using explanation data outside the region where the fitted model has support and then over-interpreting extrapolation.
How to check the result
- Repeat the explanation on held-out/subsampled data and check whether the important pattern is stable.
- Compare at least one alternative explanation or direct prediction perturbation for a small case.
- State what the explanation depends on—model, background data, feature dependence and prediction point—before drawing a decision conclusion.
Demonstrate understanding
Try this:
Build a tiny, inspectable example of PDP and ICE. First start with the audience and decision. Then choose an explanation method that matches model type and question (global vs local). Write the expected result before running it, and explain one condition that would make the result misleading or invalid.
Pick one prediction or a small held-out sample. Change one feature/condition deliberately and compare the model response with the explanation you expected.
Check reasoning, not memorisation
Before trusting a result from PDP and ICE, which check provides the strongest evidence that you understand and applied it correctly?
Keep the important distinctions visible
Step 1Start with the audience and decision.Step 2Choose an explanation method that matches model type and question (global vs local).Step 3Use held-out data when measuring importance that depends on predictive performance.Step 4Check correlated features because importance can be shared or displaced.What to remember
- PDP and ICE is part of model interpretation and communication. Explanations describe model behaviour under specific assumptions; they are not automatically causal statements about the real world.
- Start with the audience and decision.
- Presenting feature importance or attribution as causality.
- Repeat the explanation on held-out/subsampled data and check whether the important pattern is stable.