Examining Multilevel Multiple Imputation Methods for Propensity Score Analysis With a Continuous Treatment

• In Event: 67.040 - Missing Data Considerations, Solutions, and Issues

Abstract

Occurrence of missing values in multiple levels of multilevel data triggered the search for finding adequate methods to deal with missingness in complex structures. Occurrence of missing values may deteriorate the performance of the Propensity Score Analysis (PSA) to balance the treatment groups. Thus, missing values need to be handled prior to PSA. The aim of this study is to investigate the performance of multilevel multiple imputation (MMI) methods to deal with missing values in level-1 and level-2 for multilevel PSA with a continuous treatment. In addition, this study investigated to what extent MMIs perform differently from single level imputation to deal with missing data for PSA with a continuous treatment.

Authors

• Dee Duygu Cetin-Berber, University of Florida

• Walter L. Leite, University of Florida