Extra colloquium Mathematics: Hassan Pazira, MSc

Title: High-dimensional variable selection for GLMs

Abstract:

The focus of this talk is on the statistical numerical approaches to fit sparse high dimensional features with GLMs. I will describe on the selection of explanatory variables that may affect a univariate outcome. The outcome has a probability distribution that falls in the class of the exponential dispersion family. The approach that is explored is the differential geometry least angle regression (dgLARS) that is developed for generalized linear models. The dgLARS approach is compared to alternative methods for variable selection in generalized linear models. The numerical procedures of dgLARS is improved for the general setting, and is referred to as the extended dgLARS. Moreover, we investigate how well the dispersion parameter of the family of exponential distributions can be estimated. By using the simulation studies, it is shown that the improved and developed numerical procedures are fast and accurate in the estimation of parameters. In the end, an application to diabetes data is presented.

Colloquium coordinators are Prof.dr.A.J. van der Schaft ( a.j.van.der.schaft rug.nl ) and Dr. A.V. Kiselev (e-mail: a.v.kiselev rug.nl )

http://www.rug.nl/research/jbi/news/colloquia/mathematics-colloquia/