Methodology session: Methods for multilevel modeling and design effects in Add Health

Citation

Christ, Sharon L. (2016). Methodology session: Methods for multilevel modeling and design effects in Add Health. 2016 Add Health Users Conference. Bethesda, MD.

Abstract

This presentation provides an overview of two approaches to modeling Add Health data while correcting for the complex sample design features of the Add Health study. Multilevel and marginal modeling approaches will be discussed, including longitudinal modeling within these frameworks. Special focus will be on how to properly adjust model estimates for the complex sample design of Add Health, including applying the correct sampling weights and correcting for the non-independence (clustering) of observations. Treatment of missing data will also be touched upon due to its relationship to sample weight selection. Incorporation of contextual variables available in the Wave I, II, III, and IV Contextual files and the Wave I and III ONE files within these model types will be addressed. The pros and cons of the different modeling approaches and estimation methods will be considered in light of the Add Health design and the available weight variables.

URL

https://addhealth.cpc.unc.edu/wp-content/uploads/docs/events/2016%20Add%20Health%20Users%20Conference%20Abstracts_2016_06_16.pdf

Reference Type

Conference proceeding

Book Title

2016 Add Health Users Conference

Author(s)

Christ, Sharon L.

Year Published

2016

City of Publication

Bethesda, MD

Reference ID

6388