Abstract:
Count data are usually fitted through Poisson models or simple two-parameter models such as the Negative Binomial distribution. Whilst a range of numeric methods and statistical tests do exist for assessing or comparing model fit in this context, diagrammatic methods are few. We present here a diagnostic plot, which we refer to as a `Quantile Band plot', that may be used to visually assess the suitability of a given count regression model. In the case of diagnosed model inadequacy, the plot has the unique feature of conveying precise information on the character of the violation, hence pointing the data analyst towards a potentially better model choice. The graphical tool is illustrated by means of data examples from biological dosimetry, where counts of chromosome aberrations or protein foci from blood samples are used to infer the contracted radiation dose following exposure to ionizing radiation (joint work with Paul Wilson).