Loading...
Please wait, while we are loading the content...
Similar Documents
Why factor analysis is often the incorrect model for analyzing bipolar concepts, and what model to use instead (1994)
| Content Provider | CiteSeerX |
|---|---|
| Author | Schuur, T. H. Van Kiers, Henk A. L. |
| Abstract | Factor analysis of data that conform to the unfold-ing model often results in an extra factor. This artificial extra factor is particularly important when data that conform to a bipolar unidimensional unfolding scale are factor analyzed. One bipolar dimension is ex-pected, but two factors are found and often are inter-preted as two unrelated dimensions. Although this extra factor phenomenon was pointed out in the early 1960s, it still is not widely recognized. The extra factor phenomenon in the unidimensional case is reviewed here. A numerical illustration is provided, and a num-ber of diagnostics that can be used to determine whether data conform to the unidimensional unfolding model better than to the factor model are discussed. These diagnostics then are applied to an empirical example. Index terms: factor analysis, factor inter-pretation problems, rating scales, unfolding diagnos-tics, unfolding model. A technique widely used for data collection is to ask respondents to answer questions according to rating scale categories, which may range from agree very much to disagree very much. Rating scale data often are analyzed with factor analysis, which iden-tifies variables that can be attributed to the same latent factor and provides the researcher with a mea-surement value-a factor score-for each individual on each of the factors. The use of factor analysis has sometimes led to unexpected results. In particular, indicators that were meant to operationalize one bipolar dimension have been found, surprisingly, to represent two inde-pendent, unipolar factors. For example, in studies of mood state, self-descriptive adjectives like &dquo;happy&dquo; and &dquo;sad&dquo; have not all loaded highly (but with opposite signs) on the same factor, but instead have loaded highly and positively on only one of two independent, unipolar factors (e.g., Diener & |
| File Format | |
| Journal | Applied Psychological Measurement |
| Language | English |
| Publisher Date | 1994-01-01 |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
