Abstract

Testing the number of components in finite mixture models has been a long-standing challenging problem. We consider likelihood-based testing of the number of components in a finite mixture model. We develop a likelihood-based test of the null hypothesis of m components against the alternative hypothesis of m+1 components employing the EM approach of Li, Chen and Marriott (2009). The proposed EM test statistic has an easy-to-use chi-square mixture asymptotic distribution for any m. Simulation results show that the proposed test has a good finite sample size and power. A dataset from medical study is analyzed to illustrate the application of the proposed test.