Computational Mathematics & Statistics Seminar by Yiou Li: An Efficient Design of Pilot Experiments for Generalized Linear Models

Speaker:

Yiou Li, Assistant Professor, Mathematical Sciences Department, DePaul University

Title:

An Efficient Design of Pilot Experiments for Generalized
Linear Models

Abstract:

The experimental design for a generalized linear model (GLM) is interesting but challenging, since the design
often depends on model specification including the link function, the linear predictor, and the unknown regression
coefficients. However, in the pilot experiment, there is little information available regarding the model specification of
GLMs. Surprisingly, there are very limited research on investigating the design problem of pilot experiments for
GLMs. In this work, we establish some theoretical understanding on the design efficiency of pilot experiments for GLMs.
Consequently, we propose to use the low-discrepancy design for pilot experiments for GLMs. The performance of the proposed design is examined through several numerical examples.

 

 

Computational Mathematics & Statistics