When Is It Worth-While to Search for Additional Predictors?

Author(s):

Gulliksen, Harold O.

Publication Year:

1951

Report Number:

RM-51-11

Source:

ETS Research Memorandum

Document Type:

Report

Page Count:

4

Subject/Key Words:

Correlation, Factor Analysis, Predictor Variables, Research Design, Test Construction

Abstract

A preliminary consideration of test construction techniques gives four (at least superficially different) methods for determining whether or not there is something remaining in a criterion to be predicted above and beyond that which is accounted for by any particular test battery. These are: 1) correction of attenuation; 2) partial correlation; 3) factor analysis; and 4) compound symmetry. The uses and interpretation of results of each method are discussed separately. It is also suggested that, before embarking on certain long-term research projects, it would be worthwhile to devise a method "for determining in any situation whether or not one still has something remaining in the criterion over and above that which is already represented in the tests."

Read More

• Request Copy (specify title and report number, if any)