In my work with the ITC Project, I have developed some technical documentation describing the practical implementation of genearlized estimating equation (GEE) regression models using complex survey data as well as methods that account for time-in-sample. GEE models can be estimated for the longitudinal analysis of complex survey data using both SAS and SUDAAN. Such models can be used to estimated temporal trends as well as differences between groups over time (analogous to a difference-in-differences approach).

Time-in-sample is a phenomenon encountered in longitudinal surveys and arises when responses to survey questions provided by participants vary accourding to the number of times they have participated in the survey. Regression modelling makes it possible to model the relationship between an explanatory variable and an outcome while controlling for time-in-sample effects.