Activity Number:
|
658
- Regression, Selection and Complex Data
|
Type:
|
Contributed
|
Date/Time:
|
Thursday, August 1, 2019 : 10:30 AM to 12:20 PM
|
Sponsor:
|
International Indian Statistical Association
|
Abstract #304366
|
|
Title:
|
Log-Linear Model Selection and Inference for Contingency Tables
|
Author(s):
|
Arnab Chowdhury* and Subir Ghosh
|
Companies:
|
BRI, City of Hope, Duarte, CA and University of California, Riverside
|
Keywords:
|
Contingency tables;
Log-linear models;
Unsaturated models;
Saturated model;
Model selection;
Criterion functions
|
Abstract:
|
We consider several classes of unsaturated log-linear models for a contingency table with m categorical variables each at 2 levels. We search for the best fitted unsaturated model to describe the data within each class using the different criterion functions. We propose a new criterion function for the model selection based on their orthogonal and standard extensions to saturated log-linear model. Various properties of the orthogonal extensions are established. Simulations are performed to identify situations where the proposed new criterion function performs better than the other criterion functions available in the literature.
|
Authors who are presenting talks have a * after their name.