Hypothesis Testing in Mixture Regression Models ( Mathematical Details )

As a technical supplement to Zhu and Zhang (2004), we give detailed information on how to establish asymptotic theory for both maximum likelihood estimate and maximum modified likelihood estimate in mixture regression models. Under specific and reasonable conditions, we show that the optimal convergence rate of n− 1 4 for estimating the mixing distribution is achievable for both the maximum likelihood and maximum modified likelihood estimates. We also derive the asymptotic distributions of the two log-likelihood ratio testing statistics for testing homogeneity.