:package: :game_die: R/txshift: Efficient Estimation of the Causal Effects of Stochastic Interventions, with Corrections for Outcome-Dependent Sampling
This PR is adds support for censoring in the loss to follow-up sense, by adding an additional inverse probability of censoring (IPC) weighting procedure for both the one-step and TML estimators. In each case, this involves re-weighting the full-data EIF by the IPC weights. This necessitates several changes throughout, including the addition of an argument C_cens for loss to follow-up censoring and renaming of previous argument to clarify inverse probability of censoring v. sampling weights (e.g., C to C_samp and ipcw_fit_args to samp_fit_args).
This PR is adds support for censoring in the loss to follow-up sense, by adding an additional inverse probability of censoring (IPC) weighting procedure for both the one-step and TML estimators. In each case, this involves re-weighting the full-data EIF by the IPC weights. This necessitates several changes throughout, including the addition of an argument
C_censfor loss to follow-up censoring and renaming of previous argument to clarify inverse probability of censoring v. sampling weights (e.g.,CtoC_sampandipcw_fit_argstosamp_fit_args).