zhangly811/MvDeconfounder: Deconfounder

Deconfounder is a causal inference model for estimating the treatment effects of medications with eletronic health records. We aim to assess the treatment effects of multiple medications on outcomes of interest (e.g., lab measurements) using the Deconfounder. Deconfounder identifies the causal medications that have either direct effect or adverse effect on each clinical measurement. The inputs to Deconfounder are medication records and pre-treatment and post-treatment measurement values. Deconfounder fits a probabilistic factor model (e.g., poisson factoriztion or deep exponential family) to the medication records to construct substitute confounders and adjusts for substitute confounders in the outcome model for assessing the causal effects of medications..

install.packages("remotes")
remotes::install_github("zhangly811/MvDeconfounder")