The Use of Cumulant Methods for Distinction of Diagnostic Signals with Gamma Distribution

摘要

Based on the cumulant coefficients - skewness and excess kurtosis, the problem of distinction of diagnostic signals having gamma distributions with different values of the shape parameter is considered. For the given probabilities of diagnostic errors, the threshold values and minimum sample sizes were obtained for two cases - when using one of the coefficients and a pair of coefficients. The reliability of the distinction of diagnostic signals is confirmed by the results of computer simulation of signal realizations and estimates of their cumulant coefficients.