NotesFAQContact Us
Collection
Advanced
Search Tips
ERIC Number: ED450150
Record Type: Non-Journal
Publication Date: 2001-Feb
Pages: 34
Abstractor: N/A
Reference Count: N/A
ISBN: N/A
ISSN: N/A
Factor Scores, Structure Coefficients, and Communality Coefficients: It's All One General Linear Model.
Marsh, S. Neil
This paper explains the meaning and use of three important factor analytic statistics: factor scores, factor structure coefficients, and communality coefficients. For the discussion, 301 observations of junior high school students 11 measured variables from a previous study are analyzed. While factors provide the researcher with general information, the factor scores are weighted combinations of scores on a series of measured variables. Four methods of calculating factor scores are discussed. These are: (1) the regression method; (2) the Bartlett method (M. Bartlett, 1937), which uses least squares procedures to minimize the sum of squares of the other unique factors over the range of variables; (3) the Anderson-Rubin method (T. Anderson and H. Rubin, 1956), which is similar to the Bartlett method with an added condition requiring that factors scores be orthogonal; and (4) the Thompson method (B. Thompson, 1993). Calculating factor structure coefficients and communality coefficients is described, and what each brings to the analysis is outlined. In the example studied, the interpretable results were 5 extracted factors and 301 factor scores for each of the factors, 55 structure coefficients, and 11 communality coefficients. The combined evaluation of these statistics enables the researcher to formulate interpretations more accurately. An attachment contains the syntax reference for the analysis. (Contains 9 tables and 23 references.) (SLD)
Publication Type: Reports - Descriptive; Speeches/Meeting Papers
Education Level: N/A
Audience: N/A
Language: English
Sponsor: N/A
Authoring Institution: N/A