ERIC Number: ED317616
Record Type: Non-Journal
Publication Date: 1990-Jan-19
Reference Count: N/A
Backward Variable Elimination Canonical Correlation and Canonical Cross-Validation.
This paper suggests that multivariate analysis techniques are very important in educational research, and that one multivariate technique--canonical correlation analysis--may be particularly useful. The logic of canonical analysis is explained. It is suggested that a backward variable elimination strategy can make the method even more powerful, by yielding more parsimonious results and greater power against Type II error. Backward variable elimination canonical correlation analysis uses communality coefficients to eliminate variables with relatively small explanatory power. It is also suggested that cross-validation procedures should be implemented to augment interpretation. These analyses are illustrated using a small heuristic data set. Five data tables are included. (Author/TJH)
Sandra Eason, University of New Orleans, Lakefront, New Orleans, LA 70148.
Publication Type: Reports - Research; Speeches/Meeting Papers
Education Level: N/A
Authoring Institution: N/A