Research row-level statistics e.g. Shepley
hanneoberman opened this issue · comments
The Shapley value is a model agnostic measure that indicates the contribution of a particular variable to a case's prediction. It requires a sampling procedure for each case for each feature. Because they indicate row-specific contribution of a feature, an application of Shapley values to imputation models is not (yet) obvious. More information is in e.g. this gitbook.
Thank you!