Align Matching to WeightEstimator interface
#53
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Will allow Matching to be used as weight_model in `WeightedSurvival`, and allow integration with the `evaluation` suite. include tests. Signed-off-by: Ehud-Karavani <ehud.karavani@ibm.com>
Signed-off-by: Ehud-Karavani <ehud.karavani@ibm.com>
Signed-off-by: Ehud-Karavani <ehud.karavani@ibm.com>
Predicting the treatment label is only defined when you have a propensity model, not a general weighting model. Treatment label prediction is used for some of the `evaluation` metrics, but those metrics are not well-defined for weight models. Since the weight models can only be evaluated on their covariate balance (currently), then it no longer need to use the compounded Result object `WeightEvaluatorScores` and can return a covariate-balancing DataFrame (similar to how `OutcomePredictions` only returns a DataFrame with the scores). Signed-off-by: Ehud-Karavani <ehud.karavani@ibm.com>
Now that the compounded evaluation (prediction metrics and covariate balancing) is only used by the `PropensityPredictions` (and not `WeightPredictions`), the object obtaining those scores should reflect it's only relevant for propensity (treatment-prediction) models. Otherwise, it might cause confusion. Signed-off-by: Ehud-Karavani <ehud.karavani@ibm.com>
Signed-off-by: Ehud-Karavani <ehud.karavani@ibm.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Matching can be naturally presented as a weight model (in the most basic scenario, weighting samples as zero and ones for whether they were matched or excluded).
Current implementation of
Matchingalso supports this with thematching_to_weights()function.However, the API did not align with
WeightEstimator'scompute_weights(X, a)interface.This PR fixes it and realigns
MatchingtoWeightEstimator.This allows
Matchingto be evaluated usingevaluate(for example,plot_covariate_balance()) and allows it to be plugged into other models using weight-models such asWeightedSurvival(which can now generate, for example, matched Kaplan-Meier curves).This PR also includes an adjustment to the
evaluationmodule to allowMatchingto be used - moving the treatment label prediction (used for classification metrics) from theWeightPredictionstoPropensityPredictions