Skip to contents

Selecting covariates

Post-double-selection LASSO: the outcome within each treatment arm, and each arm indicator on the candidates.

lasso_select_covariates()
Select covariates by post-double-selection LASSO
lasso_select_one()
Select covariates that predict one variable

Adjusted estimators

Fit the selected set. Each falls back to a simpler specification when selection returns nothing or the adjusted fit is degenerate.

lm_lin_lasso()
Lin estimator on LASSO-selected covariates
lm_robust_lasso()
Additive robust regression on LASSO-selected covariates
lm_moderator_lasso()
Treatment-by-moderator model with LASSO-selected covariates

Attrition

Whether treatment predicts who is missing, allowing the pattern to differ across covariates.

check_attrition_lasso()
Test for differential attrition with LASSO-selected covariates

What actually ran

A fallback changes what the estimate is, so it is recorded rather than hidden. Which accessor you need depends on whether you still hold the fit.

adjustment() selected_covariates() fallback_reason()
What adjustment did a fit actually use?
fallback_summary()
Summarize what a collection of fits actually did
fallback_log() reset_fallback_log()
Record of what every fit did

Package

estimatrTools estimatrTools-package
estimatrTools: Data-Driven Covariate Adjustment for the 'estimatr' Estimators