NONPARAMETRIC REGRESSION-ANALYSIS OF DATA FROM THE AMES MUTAGENICITY ASSAY

The Ames assay has received widespread attention from statisticians because of its popularity and importance to risk assessment. However, investigators have yet to routinely apply modern regression methods that have been available for more than a decade. We study yet another approach, the application of nonparametric regression techniques, not as the ultimate solution but rather as a framework within which to address some of the shortcomings of other methods. But nonparametric regression is itself prone to difficulties when applied to Ames assay data, as we show through the use of two examples and some simulation studies. We argue that there remains a great need for further development of statistical methods suitable to the Ames assay. It is hoped that such work can be stimulated and guided by greater collaboration between statisticians and laboratory investigators.