ABSTRACT
Statistical methods provide an essential tool set for use across the field of toxicology, and although many of these have been discussed elsewhere (Maines et al., 1999) and will be discussed here, there are also methods that have particular application in the area of immunotoxicology. Although the primary focus of most of these methods is hypothesis testing, that is, the determination of whether two or more groups are different, there are two other functions that statistical methods serve, either separate from hypothesis testing or as an adjunct to hypothesis testing. The first of these functions is model fitting, which allows researchers to describe a response variable or a set of response variables in terms of another set of variables. The fitting of a variable that depends on time, such as body weight, by a polynomial function, is an example of model fitting. The second function is that of data reduction. The aim of data reduction is to reduce the dimensionality of the data with a minimum of loss in information. The familiar descriptive statistics, means, standard deviations, medians, and so on constitute a subset of this function, but model fitting can also be used to reduce dimensionality. The above example of model fitting, that is, body weight as a function of time, is also an example of data reduction. A large data set consisting of the daily weights for a set of animal subjects can be replaced by the much smaller set of parameters that define the fitted curves for each animal, usually consisting of only a few parameters per subject. This smaller set of parameters can be subsequently used for hypothesis testing, and this will be illustrated with examples of selected immunotoxicologic assays.
