We are developing and applying new methodologies to research in epidemiology. This year, work haa progressed in six areas: (1) we have developed flexible analytical methods for estimating combined effects of multiple exposures, when analyzing data from case-control studies; (2) we have clarified the fundamental notion of confounding in epidemiology, specifying that factors caused in part by an exposure under study should not routinely be adjusted for; (3) we have developed parametric extension of models for the probability of conception as related to timing of intercourse (in relation to ovulation); (4) we have shown that, under assumed independence of a genetic factor and an environmental exposure, the interaction parameter for their joint effect can be efficiently estimated based on cases only, without a control group; (5) we have begun to develop a maximum likelihood approach for estimating the fraction of birthweights that are from the """"""""residual"""""""" low end of the distribution based on population-based vital statistics records; and (6) we have identified important sources of bias which can distort findings in studies of pregnancy loss or studies of fertility based on time to conception.
