Model-Based Inference on average causal effect in clustered data

• Activity Number: 660
• Type: Contributed
• Date/Time: Thursday, August 4, 2011 : 10:30 AM to 12:20 PM
• Sponsor: Section on Government Statistics
• Author(s): Meng Wu Wu*+ and Recai M. Yucel
• Companies: The State University of New York at Albany and The State University of New York at Albany
• Keywords: causual effect ; clustered data ; mixed model ; ACE ; variance ; dual-model

Abstract

We study causal inference using potentially observable framework in clustered data(e.g. intervention studies on students nested within schools). We employ mixed-effects models to derive inference on average causal effect (ACE). Our methods apply the concept of potential outcomes in Rubin's model and extend Schafer's method of estimating the variance of ACE (Rubin, 2004; Schafer, 2008). Particularly, we develop three methods. The first one is based on linear mixed-effects model in which cluster effect is incorporated by random intercepts. The other two methods are based on dual-model strategy which reduces the confounding effects in non-randomized studies by adjusting the residuals in the linear mixed model using the inverse propensity scores. The two dual- model methods estimate the propensity scores with and without incorporating clustering. A simulation study is presented to assess the performance of the methods.