NotesFAQContact Us
Search Tips
ERIC Number: ED310152
Record Type: Non-Journal
Publication Date: 1989-Mar
Pages: 22
Abstractor: N/A
Reference Count: N/A
Two Approaches to Bootstrap Estimation in Canonical Analysis.
Schmitt, Dorren Rafael
Generalizability or invariance procedures have been known for over three decades. Through the years, these procedures have not been widely discussed or employed. One reason for the lack of use is that most of the articles on invariance procedures have been mathematically oriented. The mathematical orientation of research articles and the lack of mathematical background on the part of educational researchers have inhibited the use of generalizability procedures. An increasing number of researchers is recognizing that significance testing does not inform the researcher regarding the replicability of results and are turning to cross-validation, jackknife, and bootstrap methods as part of their analyses. Bootstrap methods seem particularly useful. This paper presents two approaches to bootstrapping in the canonical case. The first method employs bootstrap estimates of the correlation matrix from which conventional canonical functions are then derived; these correlations can be computed using software such as that provided by C. E. Lunneborg (1987). The second approach employs bootstrap estimates of the canonical correlation coefficients; these coefficients are squared and the two extreme coefficients become the upper and lower bounds for the actual canonical coefficient. The data set used was composed of 21 college students who volunteered to take the Nelson Denny Reading Test. A 33-item list of references and seven data tables are included. (Author/TJH)
Publication Type: Reports - Evaluative; Speeches/Meeting Papers
Education Level: N/A
Audience: N/A
Language: English
Sponsor: N/A
Authoring Institution: N/A