Accession Number : ADA186584

Title :   Strong Consistency of Certain Information Theoretic Criteria for Model Selection in Calibration, Discriminant Analysis and Canonical Correlation Analysis.

Descriptive Note : Technical rept.,

Corporate Author : PITTSBURGH UNIV PA CENTER FOR MULTIVARIATE ANALYSIS

Personal Author(s) : Nishii, R ; Bai, Z D ; Krishnaiah, P R

PDF Url : ADA186584

Report Date : Dec 1986

Pagination or Media Count : 19

Abstract : This paper shows that the criteria for model selection based upon efficient detection (ED) criterion are consistent for certain problems in multivariate calibration, discriminant analysis and canonical correlation analysis. These results will be proved under mild conditions on the underlying distribution.

Descriptors :   *INFORMATION THEORY, *MULTIVARIATE ANALYSIS, CALIBRATION, CORRELATION, DETECTION, DISCRIMINATE ANALYSIS, SELECTION, REGRESSION ANALYSIS, MATHEMATICAL MODELS

Subject Categories : Statistics and Probability

Distribution Statement : APPROVED FOR PUBLIC RELEASE