Parametric Statistical Methods and Sample-Size Considerations for Dominant Lethal Experiments: The Use of Clustering to Achieve Approximate Normality

摘要

While the Dominant Lethal (DL) assay is often used to assess the mutagenic potential of suspect agents, questions remain concerning its sensitivity, the appropriate way to anlayze resultant data and the necessary study size. Many questions surrounding the statistical analysis methods relate to a specification of the basic independent unit of expeirmentation and analysis, the particular type of study design used and the distributional properties of the study variables. These statistical analysis issues must be resolved before adequate attention can be given to the related questions of experimental sensitivity and sample size. This report deals with the development of a statistical analysis methodology which incorporates a particular type of data clustering scheme to achieve desirables so that analysis of variance (AOV) and related parametric testing procedures can be used. Further, the differences between dose groups which can be detected using 20, 10, and 5 cluster units of 6 males and 18 females each are estimated for each variable according to the size of selected Type I and II errors. While these procedures were developed in part using data from two DL studies on the effects of differing doses of n-Butyl Glycidyl Ether in mice, the clustering scheme suggested can be used with other AOV and regression procedures corresponding to the specific study desgin employed. (ERA citation 08:013179)