The focus of this project is the development of appropriate statistical methodologies for analysis of molecular genetic data from a variety of biological systems. Investigations continued into the modeling and analysis of data from developmental toxicity experiments, with specific attention directed at the dominant lethal assay in the male mouse. Of interest is identification of the sampling distribution of various endpoints from such experiments. A large control database was assembled, from which sampling characteristics of the endpoints of interest were determined. Statistical methods for analyzing data conforming to these characteristics are of the form of distribution free analyses, and require minimal assumptions on the distribution of the response. Various forms of such analyses were compared to identify which of the methods provide optimal statistical characteristics, such as minimal type I error, and maximal power. Permutation trend tests were seen to be quite stable, and were recommended for use. Investigations also continued into the examination and analysis of genetic susceptibility in animal and human subjects. Statistical models and methods were identified for assessing gene-environment interactions; these include logistic regression, complementary log regression, and other generalized linear models.
