I recently discovered WeightIt R package and was very happy with its functionality and performance. I “delegated” my code computing IPTW to WeightIt and it was faster while producing the same results, as expected.
In this post, I will play around with simulated data. The things I’ll be doing:
Simulating my own dataset with null associations between two different exposures (x1 and x2) and outcomes y1 and y2 for each of exposures (4 exposure-outcome pairs) Computing propensity scores (PS) for each exposure, trimming non-overlapping areas of PS distribution between exposed and unexposed Running several logistic regression models Crude Conventionally adjusted Adjusted with standardized mortality ratio (SMR) weighting using PS Calculating how biased the the estimates are compared with the true (null) effect Data simulation First, I simulate the data for 10 confounders c1-c10, 2 exposures x1 and x2 (with 7% and 20% prevalences, respectively), 2 outcomes (y1 and y2), two exposure predictors c11-c12, and 2 predictors of the outcome c13-c14.
In this post, I explore parametric g-formula fitting in the causal survival analysis context. I use the machinery of the tidyverse throughout the post and finish with plotting the 95% confidence band around the g-formula fitted survival curve for smokers vs non-smokers (see Chapter 17, Hernán MA, Robins JM (2020).
In this post, I have a look inside the Chapter 17 on Causal Survival Analysis of the “Causal Inference: What If” book by M. Hernan and J. Robins. I explore IPTW fitting following the chapter’s narrative and use the machinery of the tidyverse throughout.