Analyses

Reproducible analysis documents

Each analysis below is a reproducible document with the code and output in one place. They differ in what the outcome data allow.

HTE analysis pipeline

For outcomes defined for every patient. Conditional average treatment effects estimated with BART, subgroups discovered by fitting a tree to those estimates, sensitivity to the prior and to other estimators, heterogeneity by prognostic score, and randomization-based bounds on the distribution of individual effects.

Heterogeneous survivor average causal effects

For outcomes truncated by a terminal event, where the outcome is undefined or unobserved for participants who die. Principal stratification defines the effect among always-survivors, and BART estimates the conditional average causal effects on that stratum.

The tutorial walks through running the HTE analysis pipeline.

If you would like help applying any of this to a trial of your own, please get in touch.