A test based on the exact probability distribution of the chi 2 statistic--incorporation into the MASC method.

The MASC method (Clerget-Darpoux et al. 1988) to test genetic models in HLA associated disease makes use of a test of goodness-of-fit on categorized data. This test is composed of a sum of independent chi 2s, which are applicable only when expectations in each category are not too small. The commonly used approach to solve this problem is to combine classes in order to raise expectations. This, however, means losing some of the information contained in the data and may decrease greatly the power of discrimination between models. A test based on the exact probability distribution of a test statistic, namely the chi 2, has been implemented to avoid this dilemma.