Workshops

Multiple Imputation Analysis for Longitudinal Surveys

May 23, 2024, 2-5PM ET

This is a zoom-only workshop. Please contact [email protected] for the zoom information.

Trivellore Raghunathan

Abstract: Longitudinal or panel surveys are useful to investigate individual level changes over time. However, attrition or drop-out of individuals across waves of data collection may introduce bias in the inferences, if the mechanism leading to missing data is ignored. In addition, the surveys may also be subject to item-missing data due to nonresponse to individual questions within each wave. Multiple imputation provides a framework for incorporating missing data mechanism in constructing inferences on both time-specific and time varying estimands of interest. Several approaches for performing multiple imputation and subsequent analysis of multiply imputed data have become available in many different software packages.

This workshop will introduce a conceptual understanding of missing data mechanisms, its impact on the analysis and then discuss multiple imputation for missing data using case studies and simulated data sets. The workshop will also consider options for incorporating complex design features in the imputation and in the analysis.

Speaker Bio: Trivellore Raghunathan is Research Professor in the Michigan Program in Survey and Data Science at the Institute for Social Research and Professor of Biostatistics in the School of Public Health. He is a former chair of the Biostatistics department and a former Director of Survey Research Center. He has written two books, “Missing Data Analysis in Practice” and “Multiple Imputation in Practice”, and has developed a software IVEware for performing multiple imputation and complex survey data analysis.