ABSTRACT
This chapter introduces a technique that can be used when replicated results are available. It should therefore be seen as an extension to the PerMIA summary statistics approach. The extension takes the form of using second, rather than first order response surface models to predict the mean and variance in response. The dual response surface methodology is designed to overcome some of the main shortcomings of the generalised linear model. One of the main reasons for using a PerMIA statistic is that such a transformation enables a distinction to be made between a control and a signal factor. There is no guarantee that a PerMIA statistic that leads to separate control and signal factors can be found. It will often be the case that those factors that influence the mean will also influence the variability that is independent of the mean.
