This project encompasses the development of new and the evaluation of both old and new statistical methods appropriate for the types of data arising in epidemiologic research. This year, statistical methods for the analysis of human fertility have been studied. New methods have been developed and evaluated. In the past, fertility has often been studied in a crude fashion by techniques like counting the number of live births a couple has or by measuring the intervals between births. These sorts of techniques are unreliable, because in a country where effective birth control is widely available, the influence of these contraceptives can be so strong as to mask any possible effect of environmental toxins on fertility. Thus, in cooperation with the Epidemiology Branch, we are continuing efforts to develop methods based on analysis of the number of non-contracepting menstrual cycles a couple requires to achieve pregnancy, which is a much more sensitive endpoint for study. We have previously developed a model for such data based on a beta mixture of geometric distribution. We have now investigated the use of such models in three types of designs: a prospective design where couples are followed from the start of their efforts to achieve pregnancy, a retrospective design where women who are pregnant are asked how long it took them to achieve pregnancy, and a cross-sectional design where couples who are currently trying to achieve pregnancy are asked how long they have been trying. The applicability of the beta-geometric model to each of these designs has been investigated. The influence of a possible subgroup of sterile couples in the first and third types of designs has been examined.
