Model selection for multiway categorical data using partitioning of chi-square : part II - selected examples and program documentation / A. C. Singh and N. Ferragne. : CS11-613/88-32E-PDF

"This report contains several examples on log-linear model selection for multiway categorical data taken from the literature and their analyses by the RX2 method (Singh, 1988). The RX2 method provides a fast screening procedure for model selection within the class of hierarchical models. It uses ranking of all parameters in a decreasing order of importance as measured by P-values. RX2 provides an alternative to the ranking method based on the well- 2 known partitioning of the likelihood ratio test statistic, G which can be quite expensive computationally. Use of RX2 for the problem of polytomous logit model selection is also considered. Detailed computer program codes (in SAS-IML) are documented in the appendix"--Abstract.

Lien permanent pour cette publication :
publications.gc.ca/pub?id=9.837456&sl=1

• Format MARC XML
• Format MARC HTML

Demander des formats alternatifs