EDF Statistics for Testing for the Gamma Distribution, with Applications.

Several years ago the authors found percentage points for EDF statistics for the Gamma distribution with known shape parameter but unknown scale parameter the origin of the distribution was assumed to be zero. The gamma or equivalently the chi-squared distribution often occurs as a possible model for observations or as the distribution of some derived statistics. For example, in lifetime or survival studies the gamma distribution can be proposed as the distribution of lifetime or some function of lifetime. The chi-squared distribution occurs as the distribution for sums of squares in ANOVA tables and also in time series analysis, in the study of periodograms, and in multivariate analysis, as the distribution of squared radii. Despite the many uses of the gamma distribution, the application of the tables might seem to be limited, because the shape parameter must be known. In the earlier reports, are an illustration of the use of EDF statistics to test that the sample variances of cells of an ANOVA table, with the same number of observations in each cell, all came from the same distribution. Since then, other applications have come to light and these will be presented in this report also the test procedures and the tables will be reproduced to make the report complete in itself.