Literaturnachweis - Detailanzeige
Autor/in | Halperin, Silas |
---|---|
Titel | The Incorrect Measurement of Components. |
Quelle | (1973), (10 Seiten)
PDF als Volltext |
Beigaben | Tabellen |
Dokumenttyp | gedruckt; online; Monographie |
Schlagwörter | Tagungsbericht; Correlation; Factor Analysis; Orthogonal Rotation; Research Reports; Scoring Formulas; Speeches; Statistical Analysis; Test Interpretation; Weighted Scores |
Abstract | Factor loadings, used directly or as the basis of binary values, are not appropriate as weights to produce component scores from a rotated solution. A series of examples showing the results of an incorrect measurement of components is given. Several correlation matrices were taken from books on factor analysis and multivariate analysis. Each matrix was submitted to a principal components decomposition, and those vectors whose roots were greater than 1.0 were then rotated according to the normalized varimax criterion. Following this, several matrices were calculated. A clear pattern resulted. Components supposedly orthogonal to each other often come out highly correlated. In addition, factor loadings that are small in the rotated principal components structure are often moderate to large in R/zh and R/zk. Thus, the researcher who interprets and names components finds that his scores have a completely different meaning than he anticipates. Any use of these scores in further analysis will result in a serious distortion of conclusions. (Author/KM) |
Erfasst von | ERIC (Education Resources Information Center), Washington, DC |
Update | 2004/1/01 |