A comparison of six statistical distributions for analysis of chromosome aberration data for radiation biodosimetry.

The Poisson distribution is the most widely recognised and commonly used distribution for cytogenetic radiation biodosimetry. However, it is recognised that, due to the complexity of radiation exposure cases, other distributions may be more properly applied. Here, the Poisson, gamma, negative binomial, beta, Neyman type-A and Hermite distributions are compared in terms of their applicability to 'real-life' radiation exposure situations. The identification of the most appropriate statistical model in each particular exposure situation more correctly characterises data. The results show that for acute, homogeneous (whole-body) exposures, the Poisson distribution can still give a good fit to the data. For localised partial-body exposures, the Neyman type-A model was found to be the most robust. Overall, no single distribution was found to be universally appropriate. A distribution-specific method of analysis of cytogenetic data is therefore recommended. Such an approach may lead potentially to more accurate biological dose estimates.