USask Mathematics & Statistics Colloquium: Pei Geng

Abstract: In Logistic regression under the case–control framework, the logarithmic ratio of the covariate densities between the case and control groups is a linear function of the regression parameters. Based on this fact, this talk introduces an integrated least-square-type framework for parameter estimation and variable selection in Logistic regression using the estimated covariate densities. Simulation study shows that the proposed approach achieves higher accuracy in parameter estimation for small to moderate samples. Moreover, it outperforms traditional regularization methods such as LASSO, Elastic Net and SCAD regarding variable selection in high dimensional settings.