Accession Number : AD0441337


Title :   A STUDY OF REDUCED RANK MODELS FOR MULTIPLE PREDICTION


Corporate Author : WASHINGTON UNIV SEATTLE


Personal Author(s) : Burket, George R


Full Text : https://apps.dtic.mil/dtic/tr/fulltext/u2/441337.pdf


Report Date : Jan 1943


Pagination or Media Count : 80


Abstract : The present study proceeds along both theoretical and empirical lines. First an attempt is made to work out some of the consequences of regression theory for reduced-rank models. Since, as noted above, there is reason to question the appropriateness of regression theory for psychological prediction problems, and empirical comparison of five reduced-rank procedures is also carried out. The methods used were predictor elimination, predictor selection, the method of approximating the intercorrelation matrix, the method of approximating the inverse, and the method using the principal-axes factors giving the highest multiple correlation. As will be seen, both the theoretical and the empirical evidence favors the method of approximating the intercorrelation matrix.


Descriptors :   *LEARNING , *RETENTION(PSYCHOLOGY) , MATHEMATICAL PREDICTION , PSYCHOLOGICAL TESTS , PSYCHOLOGY


Subject Categories : Psychology


Distribution Statement : APPROVED FOR PUBLIC RELEASE