Title: Selection of the number of components in mixture regression model
Authors: Wing Kam Fung - University of Hong Kong (Hong Kong) [presenting]
Abstract: Mixture regression model has been proven to be a useful tool in the study of heterogeneous populations arising in many fields. An important question in model building is the selection of the number of components. The majority of existing methods emphasize the goodness of fit and do not differentiate this problem with the diagnosis of other aspects of a mixture model. The classification probability instability method based on the concept of instability is introduced to select the number of components in mixture regression models. The proposed method can deal with the situation where the number of components is only one, in which the existing instability procedures may not be able to investigate. Stages are used to handle the possible multilevel structure in the components. Two variations are examined for small samples. The higher accuracy compared with some information criterion procedures are illustrated by simulations. A real data set on plasma beta-carotene concentration was analyzed. The selected mixture model offers a better interpretation about the relationship between the plasma beta-carotene concentration and dietary factors and personal characteristics.