Store estimated models for nuisance parameters

[MalteKurz]

Description

This PR implements the often requested feature to store the estimated models for nuisance parameters. To use it, call the method fit() with option store_models=True. Example:

library(DoubleML) library(mlr3) library(mlr3learners) library(data.table) set.seed(2) ml_g = lrn("regr.ranger", num.trees = 10, max.depth = 2) ml_m = ml_g$clone() obj_dml_data = make_plr_CCDDHNR2018(alpha = 0.5) dml_plr_obj = DoubleMLPLR$new(obj_dml_data, ml_g, ml_m) dml_plr_obj$fit(store_models=TRUE)

The estimated models can then be found in the attribute dml_plr_obj$models:

dml_plr_obj$models $ml_l $ml_l$d $ml_l$d[[1]] $ml_l$d[[1]][[1]] <LearnerRegrRanger:regr.ranger> * Model: ranger * Parameters: num.threads=1, num.trees=10, max.depth=2 * Packages: mlr3, mlr3learners, ranger * Predict Types: [response], se * Feature Types: logical, integer, numeric, character, factor, ordered * Properties: hotstart_backward, importance, oob_error, weights $ml_l$d[[1]][[2]] <LearnerRegrRanger:regr.ranger> * Model: ranger * Parameters: num.threads=1, num.trees=10, max.depth=2 * Packages: mlr3, mlr3learners, ranger * Predict Types: [response], se * Feature Types: logical, integer, numeric, character, factor, ordered * Properties: hotstart_backward, importance, oob_error, weights $ml_l$d[[1]][[3]] <LearnerRegrRanger:regr.ranger> * Model: ranger * Parameters: num.threads=1, num.trees=10, max.depth=2 * Packages: mlr3, mlr3learners, ranger * Predict Types: [response], se * Feature Types: logical, integer, numeric, character, factor, ordered * Properties: hotstart_backward, importance, oob_error, weights $ml_l$d[[1]][[4]] <LearnerRegrRanger:regr.ranger> * Model: ranger * Parameters: num.threads=1, num.trees=10, max.depth=2 * Packages: mlr3, mlr3learners, ranger * Predict Types: [response], se * Feature Types: logical, integer, numeric, character, factor, ordered * Properties: hotstart_backward, importance, oob_error, weights $ml_l$d[[1]][[5]] <LearnerRegrRanger:regr.ranger> * Model: ranger * Parameters: num.threads=1, num.trees=10, max.depth=2 * Packages: mlr3, mlr3learners, ranger * Predict Types: [response], se * Feature Types: logical, integer, numeric, character, factor, ordered * Properties: hotstart_backward, importance, oob_error, weights $ml_m $ml_m$d $ml_m$d[[1]] $ml_m$d[[1]][[1]] <LearnerRegrRanger:regr.ranger> * Model: ranger * Parameters: num.threads=1, num.trees=10, max.depth=2 * Packages: mlr3, mlr3learners, ranger * Predict Types: [response], se * Feature Types: logical, integer, numeric, character, factor, ordered * Properties: hotstart_backward, importance, oob_error, weights $ml_m$d[[1]][[2]] <LearnerRegrRanger:regr.ranger> * Model: ranger * Parameters: num.threads=1, num.trees=10, max.depth=2 * Packages: mlr3, mlr3learners, ranger * Predict Types: [response], se * Feature Types: logical, integer, numeric, character, factor, ordered * Properties: hotstart_backward, importance, oob_error, weights $ml_m$d[[1]][[3]] <LearnerRegrRanger:regr.ranger> * Model: ranger * Parameters: num.threads=1, num.trees=10, max.depth=2 * Packages: mlr3, mlr3learners, ranger * Predict Types: [response], se * Feature Types: logical, integer, numeric, character, factor, ordered * Properties: hotstart_backward, importance, oob_error, weights $ml_m$d[[1]][[4]] <LearnerRegrRanger:regr.ranger> * Model: ranger * Parameters: num.threads=1, num.trees=10, max.depth=2 * Packages: mlr3, mlr3learners, ranger * Predict Types: [response], se * Feature Types: logical, integer, numeric, character, factor, ordered * Properties: hotstart_backward, importance, oob_error, weights $ml_m$d[[1]][[5]] <LearnerRegrRanger:regr.ranger> * Model: ranger * Parameters: num.threads=1, num.trees=10, max.depth=2 * Packages: mlr3, mlr3learners, ranger * Predict Types: [response], se * Feature Types: logical, integer, numeric, character, factor, ordered * Properties: hotstart_backward, importance, oob_error, weights 

Note that the number of fitted models depends on the settings and the considered model. The outer named list contains one entry for each nuisance part (here ml_l and ml_m). For each nuisance part there is a named list containing an entry for each treatment variable (here only 'd'). The next inner part is a list of length n_rep (repeated cross-fitting) and then a list of length n_folds (number of folds per repeated cross fit).

PR Checklist

• The title of the pull request summarizes the changes made.
• The PR contains a detailed description of all changes and additions.
• The code passes R CMD check and all (unit) tests (see our contributing guidelines for details).
• Enhancements or new feature are equipped with unit tests.
• The changes adhere to the "mlr-style" standards (see our contributing guidelines for details).

[PhilippBach]

Thanks @MalteKurz for preparing the PR. As you indicate in the check boxes of the PR, there are still the tests missing; Either I can do them in the next days myself or I'll add my review to them later (in case you're faster 😃 ). Thanks!