Conferences

Dependent data come in many forms and may result, for example, from repeated measurements on the same individuals, interviewing family members, surveying pupils clustered within schools etc. etc. For many important research questions the dependencies as such are not of substantive interest but just a nuisance that has to be taken into account when estimating the parameters of interest. Next to autocorrelation or random effect models, marginal modeling can be used to take the dependencies in the data into account. In contrast to most applications of the other approaches, in marginal modeling, one does not need to make implicit or explicit assumptions about the nature of the dependencies. Recent developments make a full fledged ML approach possible (using the Agresti/Lang/Bergsma algorithm)

In this session papers will be discussed and contributions are invited that deal with

- applications of marginal modeling

- comparisons of estimation methods (WLS, GEE, MLE)

- comparisons between marginal models and random effect models and between
conditional (subject specific) and marginal (population averaged) approaches

- causality and marginal models