Variable selection in mixture regression models
Conference
65th ISI World Statistics Congress 2025
Format: IPS Abstract - WSC 2025
Keywords: mixture-regession, regularization, variable-selection
Session: IPS 824 - Unveiling the Power of Mixture Models in a Data-Rich World
Thursday 9 October 10:50 a.m. - 12:30 p.m. (Europe/Amsterdam)
Abstract
The selection of relevant variables through regularization has been popularly used in many studies. On the other hand, finite mixture regressions have been widely used to explore and model heterogeneous covariate effects on the response. To our best knowledge, problems of variable selection in mixture regression models has drawn little attention in the literature so far, and therefore limited methodologies available for fitting a regularized mixture regression model. In this talk, we introduce a novel algorithm for optimizing a regularized mixture regression models with different choices of penalties. Our proposed method is desirable for retaining only relevant covariates in each of the subpopulations and allowing these relevant covariates have heterogeneous effects in different subpopulations.