Abstract:
|
A normal mixture regression model is often used to characterize the relationship between the response and covariates in a population with several latent and heterogeneous subpopulations. In many practical applications, the slope difference between two regression models could be of particular interest, it can be used to compare the impact of a specific covariate over different subpopulations. In this talk, I will introduce a new method to derive a confidence interval for the slope difference in a mixture regression model. Some simulation results will be presented for showing that the proposed method can maintain the empirical coverage rate sufficiently close to the nominal level of confidence. In addition, a real-life data analysis will be given for demonstrating that the proposed method is feasible in practice.
|