Efficient Estimation of Partially Observed Clustered Data Using Multiple Imputation

• Activity Number: 499
• Type: Contributed
• Date/Time: Wednesday, August 7, 2013 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Kathryn Aloisio*+ and Nicholas J. Horton and Sonja Swanson and Alison E. Field and Nadia Micali
• Companies: Smith College and Smith College and Private and Boston Children's Hospital and UCL Institute of Child Health
• Keywords: ALSPAC Study ; eating disorders ; multiple informants ; generalized estimating equations ; multiple imputation ; missing data

Abstract

Clustered data commonly arise in social and biomedical sciences. As one example, multiple-source reports are often collected in child and adolescent psychiatric epidemiologic studies. Researchers use various informants (e.g. parent and child) to provide a holistic view of a subject's symptomatology. These studies often have missing data due to multiple stages of consent and willingness to participate. Fitzmaurice and colleagues described estimation of multiple source models using a generalized estimating equation (GEE) framework, assuming MCAR missingness. Multiple imputation is an attractive method to fit incomplete data models under the less restrictive Missing at Random (MAR) assumption. We demonstrate how to utilize multiple imputation in conjunction with a standard GEE in a study of eating disorder symptoms with parallel reports from parents and adolescents in the ALSPAC study. While point estimates were fairly similar to the GEE under MCAR, the MAR model had smaller standard errors, while requiring less stringent assumptions regarding missingness. This approach is available within general purpose statistical software, and is recommended as a principled analytic approach.