Kang Heeyeon, a master's student in the Department of Mathematics (supervised by Shin Sunyoung), received the Encouragement Award at the 2024 Korean Statistical Society Summer Academic Conference held from July 4 to 6 at Sungkyunkwan University's Humanities and Social Sciences Campus.

Paper Title: Penalized Estimation for a Finite Mixture of Multivariate Regression Models

Research Summary: In the era of big data, we often encounter high-dimensional and heterogeneous data with dependencies among variables. Therefore, it has become crucial to develop statistical models that can analyze the relationships between multiple correlated response variables and high-dimensional predictors to address the heterogeneity within big data.

In response to this need, we developed a finite mixture of regre
ssion models with multiple responses (mvFMR). The mvFMR model has the advantage of handling high-dimensional heterogeneous data while considering the dependencies among multiple response variables, allowing us to identify various subgroups and their regression relationships within the data.

We proposed the penalized maximum likelihood approach as an effective tool for variable selection and developed the EM-ADMM algorithm, which combines the EM algorithm and ADMM to obtain penalized MLE.

Through simulations, we confirmed that our method performs successfully compared to unpenalized MLE and alternative approaches that consider multiple response variables individually. We applied our method to analyze diabetes diagnosis data and shared bicycle data.