Abstract
This article suggests methods for forming linear composites which will have either optimum (maximum) criterion validity or an optimum ratio of valid to reliable variance. These methods may be of interest to researchers whose data contain substantial systematic but invalid components. These techniques are formally related to canonical correlation analysis, and the output from standard canonical correlation computer programs can be directly used in forming the composites.
Original language | English (US) |
---|---|
Pages (from-to) | 55-59 |
Number of pages | 5 |
Journal | Social Science Research |
Volume | 9 |
Issue number | 1 |
DOIs | |
State | Published - Mar 1980 |
ASJC Scopus subject areas
- Education
- Sociology and Political Science