Investigations into parametric analysis of data from in vivo micronucleus assays by comparison with non-parametric...

Data from micronucleus assays of 6 compounds were analysed by 3 non-parametric and 3 parametric methods. Two of the latter involved transformation of the data so several transformation strategies were investigated. It was concluded that the non-parametric Kolmogorov-Smirnov two-sample test was the most reliable method of analysis. None of the parametric solutions was entirely satisfactory. The Bayesian solution to the Behrens-Fisher problem of normal distributions with differing variances was an acceptable compromise after the data had been transformed by the inverse hyperbolic sine method applicable to negative binomials.