Accession Number : AD0691216

Title :   COMPUTATIONAL EFFICIENCY IN THE SELECTION OF REGRESSION VARIABLES.

Descriptive Note : Final rept. on Themis optimization research program,

Corporate Author : TEXAS A AND M UNIV COLLEGE STATION INST OF STATISTICS

Personal Author(s) : LaMotte,L. R. ; Hocking,R. R.

Report Date : JUL 1969

Pagination or Media Count : 24

Abstract : A number of criteria have been proposed for selecting the best subset or subsets of independent variables in linear regression analysis. Applying these criteria to all possible subsets is, in general, infeasible if the number of candidate variables in large. Since most criteria are monotone functions of the residual sum of squares, the problem is reduced to identifying subsets for which this quantity is small. In the report a method is described which will identify best subsets while considering only a small fraction of the possible subsets. The method is based on a branch and bound technique and will identify the best subset of each size and has the added feature that a number of nearly best subsets are also revealed.

Descriptors :   (*REGRESSION ANALYSIS, ALGORITHMS), COMPUTER PROGRAMMING, MULTIVARIATE ANALYSIS, OPTIMIZATION

Subject Categories : Statistics and Probability

Distribution Statement : APPROVED FOR PUBLIC RELEASE