Simple Multinomial Cell-Probability Learner

A small built-in learner for dml_backend = "custom" examples and tests. It estimates the four (d, post) cell probabilities with a multinomial logistic model when nnet is available and falls back to empirical cell shares if the model cannot be fit.

Usage

compdid_cell_ps(dp_train, dp_test, ps_min = 1e-05, ...)

Arguments

dp_train

Training dp object.

dp_test

Test dp object.

ps_min

Minimum cell probability before row normalization.

...

Unused.

Value

An n_test x 4 matrix of cell probabilities.