|
Activity Number:
|
390
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 1, 2007 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Social Statistics Section
|
| Abstract - #310008 |
|
Title:
|
Linear Information Models: A Prototype
|
|
Author(s):
|
Philip E. Cheng*+ and Jiun W. Liou and Michelle Liou and John Aston
|
|
Companies:
|
Academia Sinica and Academia Sinica and Academia Sinica and Academia Sinica
|
|
Address:
|
128 Academia Rd, Taipei, 115, Taiwan
|
|
Keywords:
|
Information Models ; Log-linear Models ; Model Selection ; Mutual Information
|
|
Abstract:
|
Relative entropy identities yield basic log-likelihood decompositions for categorical data analysis. It naturally develops the linear information models in contrast to the hierarchical log-linear models. A recent study by the authors (Journal of Data Science, Oct. 2006) clarified the difference in the data likelihood analysis between the two linear models. To supplement our previous study, a prototype of linear information models and a basic scheme of model selection are formulated. Empirical study with high-way contingency tables is employed to illustrate the validation of information models and the difference from the log-linear models.
|
