Data included in package

This package contains several example datasets that can be used to explore its functionality. The datasets and simulation scripts used to generate them are contained in their own dedicated branch, and described in detail below.

Simple simulated data

The table below defines the possible underlying correlation structures for the covariates in the simple simulated datasets.

• Y ~ N(mu, 1) is the continuous outcome, with the mean given in the table below.
• trt ~ Bern(p) is the binary treatment, with the mean given in the table below.
• X5 ~ N(0,1) is a covariate associated with the treatment only (i.e., an instrument).
• X6 ~ N(0,1) is a covariate associate with the outcome only (i.e., a prognostic variable).
• X1, X2, X3, X4 ~i.i.d. N(0,1) are confounders of the effect of the treatment on the outcome.
• E1, E2, and E3 ~i.i.d. Bern(0.5) are binary effect modifiers that define eight subgroups within the data.

1. Confounding and no effect modification
2. Effect modification and no confounding
3. Effect modification and confounding
4. Effect modification and confounding, with additional confounding by effect modifiers

Complex simulated data

1. Medicare data exploring the effect of stent type on revacularization rate