Estimation of Dynamic Models with Nonignorable and Nonmonotone Drop-Out

• Activity Number: 41
• Type: Contributed
• Date/Time: Sunday, August 3, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Survey Research Methods Section
• Author(s): Jongho Im*+ and Jae-Kwang Kim
• Companies: Iowa State University and Iowa State University
• Keywords: Dynamic model ; Longitudinal data ; Nonignorable nonresponse ; Nonmonotone drop-out ; Panel attrition

Abstract

We propose an alternative of weighted generalized estimating equations for incomplete longitudinal data exposed to non-ignorable and non-monotone drop-out. We introduce the generalized method of moments type estimator combining weighted generalized estimating equations of dynamic model and the mean score functions of response model which are obtained in each time point. The proposed mean score function of response model is a new approach of maximum likelihood estimation that is obtained without specifying the dynamic outcome model. It is less sensitive to failure of the assumed outcome model. We present a limited simulation study and apply our method to the analysis of Korea Work Place Survey (WPS) panel data.