CMStatistics 2015: Start Registration
View Submission - CMStatistics
B1439
Topic: Contributed on Advances in mixture modelling Title: Finite mixture regression models with concomitant variables: A simulation study Authors:  Kristyna Vankatova - Palacky University (Czech Republic) [presenting]
Eva Fiserova - Palacky University (Czech Republic)
Abstract: Finite mixture models are a popular technique for modelling unobserved heterogeneity in the population. Within the family of mixture models, mixtures of regression models have been proposed and studied as a replacement for more traditional cluster analysis and cluster-wise regression techniques. Using mixtures of regression models, the sample representing a heterogeneous population can be clustered into groups by modelling the conditional distribution of the response given the explanatory variable as a mixture. In many applications, the parameters of a mixture of linear regression models are estimated with the expectation maximization (EM) algorithm within a maximum likelihood framework. In order to characterize the different components and improve regression parameter estimates and predictions the use of concomitant variables has been proposed. The goal is to compare the performance of standard mixture of regressions and mixture of regressions with concomitant variables. A simulation study is designed to investigate this problem and is focused on the accuracy of parameter estimations and following predictions. R package flexmix is used for modelling of mixtures using the EM algorithm.