Testing for equalities of several multinomial distributions

Abstract

 In this article we have studied the testing of parameters of several multinomial distributions. The necessary conditional distribution of complete sufficient statistic of the parameter of interest is obtained by conditioning the complete sufficient statistics of nuisance parameter to be held fixed .Using this conditional distribution we have derived uniformly most powerful unbiased (UMPU) test for testing the hypothesis (i)Equality of two independent multinomial distributions and (ii) testing one of the parameters of multinomial distribution is equal to or larger than the corresponding parameters of another independent multinomial distribution. Asymptotic test have been derived for testing (i) and (ii).The application of these tests have been done to associate class in data mining.