Understanding chemically induced changes in mutational spectra can aid in deciphering mechanisms of mutagenesis. In this paper, we propose the use of statistical methods that are based upon the mutation frequency, rather than simple mutant counts which have no relationship to the mutation frequency. These methods have a number of advantages over the current standard analysis: an improved means of identifying those classes/sites of mutation which have treatment-related induction, greater sensitivity to localized differences in spectra (e.g., limited to a single base pair), one-sided tests for induction of mutations, tests of dose-response, and a framework for sample-size estimation in terms of the number of mutants to sequence. As examples, the methods are applied to data from transgenic mutation assays.