Home > Events > 2014 Seminars & Colloquia > Keri Lock Morgan - Duke University

Keri Lock Morgan - Duke University

Main Content

"Rerandomization to Improve Covariate Balance in Experiments"
23 January 2014 from 4:00 PM to 5:00 PM
201 Thomas
Add event to calendar

Abstract: Randomized experiments are the gold standard for establishing a causal effect, primarily because on average randomization balances background covariates across the treatment groups.  However, in any one experiment, it is quite possible that groups will be imbalanced on important covariates simply due to random chance.  In this talk we discuss a way in which researchers can ensure balance on important covariates by rejecting “unlucky” randomizations and rerandomizing.  The theoretical implications are discussed, including conditions for maintaining an unbiased estimate, improvement in covariate balance, and increase in precision of the estimated treatment effect.

Filed under: ,