ERIC Number: ED418125
Record Type: Non-Journal
Publication Date: 1998-Apr-13
Reference Count: N/A
Features of Published Analyses of Canonical Results.
Humphries-Wadsworth, Terresa M.
D. Wood and J. Erskine (1976) and B. Thompson (1989) provided bibliographies of roughly 130 applications of canonical correlation analysis, but the features of such reports have not been widely studied. This report examines the features of recent canonical reports, including substantive inquiries, but also measurement applications examining multivariate validity and multivariate reliability. One particular area of interest focuses on interpretation of functions as against structure coefficients. Little appears to have changed since the publication of the Wood and Erskine study. The current review of the literature yields similar results about the confusing and somewhat arbitrary use of canonical terminology. Several analyses are highlighted that illustrate why students have so much trouble understanding canonical results. In addition to using confusing terminology, many authors failed to provide all the information needed to evaluate their conclusions. Recommendations for reporting canonical results include evaluating both the squared canonical correlation coefficients and statistical significance test results to decide which canonical functions to interpret. Both the canonical function coefficients and the canonical structure coefficients should be interpreted for noteworthy functions. One should usually not try to interpret the redundancy coefficients. One must, however, examine the communality coefficients for the variables that do not contribute to the overall canonical correlation solution, and one should evaluate the generalizability of the results through statistical or empirical means. Measurement applications are outlined. (Contains 1 table and 51 references.) (SLD)
Publication Type: Information Analyses; Speeches/Meeting Papers
Education Level: N/A
Authoring Institution: N/A