Balance diagnostics for treatment effects
Treatment effects extract experiment-style causal effects from observational data. A key requirement is that our treatment-effects model explicitly or implicitly reweights the data such that the model-adjusted distribution of the covariates is comparable across treatment groups. Balance diagnostics check this.
Four diagnostics and tests are provided.
One reports, for each covariate, the model-adjusted difference in means in the treatment groups and the ratio of variances, providing a useful diagnostic.
Another graphs the model-adjusted estimated pdfs of covariates; these pdfs can be examined visually to verify that they are approximately equal.
Another does the same but uses box plots rather than smoothed pdfs.
And finally, an overidentification test is provided. It statistically tests whether the model-adjusted means of the covariates are the same between groups.