Multiple Imputation Strategy for Alameda County Study

• Activity Number: 196
• Type: Contributed
• Date/Time: Monday, August 7, 2006 : 2:00 PM to 3:50 PM
• Sponsor: Section on Survey Research Methods
• Author(s): Irina Bondarenko*+ and Trivellore E. Raghunathan
• Companies: University of Michigan and University of Michigan
• Keywords: multiple imputation ; longitudinal ; imputations ; imputation diagnostic ; missing values

Abstract

The Alameda County Study is a prospective study of a stratified random sample of members of Alameda County, California households, with data first collected in 1965, and five subsequent waves of data collection carried in 1974, 1983, 1995, and 1999. It is the longest ongoing, population-based study in the US that contains substantial amounts of information on health status and behavioral, social, socioeconomic, neighborhood, and socio-environmental risk factors. The study is subjected to numerous sources of missing data, including partial non-response, lost to follow-up, changes in questionnaires, and collection of data on a sub-sample in 1983, due to reduced funding. This paper explores multiple imputation strategy for a variety of longitudinal analyses. A number of diagnostics tests are used to check validity of the models used in imputations, and plausibility of completed datasets.