Inference in experiments conditional on observed imbalances in covariates
报告人：Per Johansson(Uppsala University)
时间：2021-12-17 16:00-17:00
地点：Zoom Meeting(865 9016 5749)
Abstract：
Double blind randomized controlled trials are traditionally seen as the gold standard for causal inferences as the difference-in-means estimator is an unbiased estimator of the average treatment effect in the experiment. The fact that this estimator is unbiased over all possible randomizations does not, however, mean that any given estimate is close to the true treatment effect. Similarly, while pre-determined covariates will be balanced between treatment and control groups on average, large imbalances may be observed in a given experiment and the researcher may therefore want to condition on such covariates using linear regression. This paper studies the theoretical properties of both the difference-in-means and OLS estimators conditional on observed differences in covariates. By deriving the statistical properties of the conditional estimators, we can establish guidance for how to deal with covariate imbalances. We study both inference with OLS, as well as with a new version of Fisher's exact test, where the randomization distribution comes from a small subset of all possible assignment vectors.
Biography：
Professor Per Johansson is from Uppsala University. He is currantly Professor at the Department of Statistics. He has been a researcher at Institute for Evaluation of Labour Market and Education Policy (IFAU).He has had a chair in micro econometric evaluation at the Department of Economics. Per won a lot of honors: such as, he is a co-opted member the Prize Committee for the Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel, and member of Statistics Sweden’s Scientific Advisory Board. Per Johansson’s research interest is in causal inference.His research deals with the analysis of micro data in general (both methodological and applied). The applied work has mainly been in labor economics (effects of income security systems on labor market behavior, program evaluation, and economics of education) but he has also done research in transport economics.
Join Zoom Meeting：https://us02web.zoom.us/j/86590165749
Meeting ID：865 9016 5749
Paper: https://arxiv.org/abs/2012.10686
