A1471
Title: C(alpha) test for number of components in finite mixture models
Authors: Junfan Tao - Kyoto University (Japan) [presenting]
Jiaying Gu - University of Toronto (Canada)
Stanislav Volgushev - University of Toronto (Canada)
Abstract: Finite mixture models are widely used in empirical applications, yet determining the number of components remains a challenging task. The aim is to propose a C(alpha) test for inference on the number of components. It is shown that it is asymptotically equivalent to the EM test but does not require tuning parameters and often yields better power performance in finite samples. Both the C(alpha) and EM tests have limiting distributions that are mixtures of chi-squares, though the weights typically require simulation. To address this, the most stringent somewhere most powerful test (MSSMP) is also proposed, as originally introduced in a past study. This test benefits from a pivotal limiting distribution. Simulation studies show that the MSSMP test possesses comparable performance to the C(alpha) test. An empirical example to demonstrate the application of these tests is concluded with.
