Dealing With Missing Data in Multilevel Contexts: A Simulation Study

• In Event: 61.087 - Poster Session 12
  In Poster Session: 61.087-4 - Multilevel Modeling: Complex Survey Data and Missing Data

Abstract

Given the prevalence of missing data when fitting multilevel models in educational research, there is a pressing need to understand the strengths and limitations of available methods that deal with this issue. Using Monte Carlo simulations, the present study examines the performance of five different methods for dealing with missing data: (1) single-level multiple imputation, (2) multilevel multiple imputation with hybrid dummy coding, (3) multilevel multiple imputation with separate imputations within clusters, (4) multilevel multiple imputation using the R package PAN, and (5) FIML. Recommendations for applied researchers will be discussed.

Authors

• Hui Jiang, The Ohio State University

• Susan Anderson Mauck, The Ohio State University - Columbus

• Mine Dogucu, The Ohio State University

• Ann A. O'Connell, The Ohio State University