A Moment-Based Test for Homogeneity in Finite Mixture Models
Wei Ning, Arjun K. Gupta, Chang Yu, Sanguo Zhang · Communication in Statistics- Theory and Methods · 2009
This research is about testing homogeneity in finite mixture models. In literature, solutions to this problem normally involve first establishing the identifiability of parameters, then testing the hypothesis whether the data come from a single distribution or a mixture of distributions. We propose a moment-based test without getting into the issue of parameter estimability. Our simulations demonstrate that the power of our test is comparable to LRT, C(α) test, and bootstrap test with the additional advantage that our test always controls the Type I error rate within the nominal α level for the normal case. We demonstrate our test on a real data set to identify potential predictive biomarkers for hospitalization in hemodialysis patients.