Abstract
Large animal toxicology studies generally contain small numbers of animals, 3–5/sex/group. Traditionally, people do not take the multivariate and multifactor aspects of the design into consideration while analyzing the parameters of interest of such studies. As a consequence, many investigators reach inappropriate and sometimes incorrect statistical conclusions from their analyses. The main purpose of this article is to show how, given the typically minimal sample sizes, one can reach more meaningful conclusions from such studies if one takes into consideration the factorial and multivariate nature of the design. The secondary purpose is to point out the need of careful examination of model specification and use of some popular statistical software. Examples are given to demonstrate the points.