|
Activity Number:
|
173
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 1, 2016 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Statistics and the Environment
|
|
Abstract #320184
|
View Presentation
|
|
Title:
|
A Penalized Quasi-Likelihood Approach for Variable Selection in High-Dimensional Spatially Correlated Binary and Count Data
|
|
Author(s):
|
Abdhi Sarkar* and Chae Young Lim and Tapabrata Maiti
|
|
Companies:
|
Michigan State University and Seoul National University and Michigan State University
|
|
Keywords:
|
Spatial ;
quasi-likelihood ;
working correlation ;
high dimension
|
|
Abstract:
|
With the unique problem of not being able to specify a likelihood with spatially correlated discrete non-normal data, in this paper we propose a variable selection technique using the quasi-likelihood approach. A generalized link function is used to describe the relationship between the first two moments of the response variable and covariates of interest. The proposed estimator is proven to estimate and select variables simultaneously. The theoretical properties of asymptotic normality and consistency are investigated under the increasing domain framework after introducing a mis-specified working correlation matrix that satisfies a strong mixing condition. In cases where the number of covariates is larger than the observed number of locations or points sampled, under the sparse model assumption we also discuss the nature of the selection consistency. A real data example is provided alongside extensive simulations to showcase the performance of our proposed method.
|
Authors who are presenting talks have a * after their name.
