Literaturnachweis - Detailanzeige
Autor/in | Samuelsen, Karen |
---|---|
Titel | Commentary on Factorial versus Typological Models: Complementary Evidence in the Model Selection Process |
Quelle | In: Measurement: Interdisciplinary Research and Perspectives, 10 (2012) 4, S.222-224 (3 Seiten)Infoseite zur Zeitschrift
PDF als Volltext |
Sprache | englisch |
Dokumenttyp | gedruckt; online; Zeitschriftenaufsatz |
ISSN | 1536-6367 |
DOI | 10.1080/15366367.2012.743402 |
Schlagwörter | Stellungnahme; Models; Statistical Analysis; Classification; Factor Structure; Selection; Evidence; Multivariate Analysis |
Abstract | The notion that there is often no clear distinction between factorial and typological models (von Davier, Naemi, & Roberts, this issue) is sound. As von Davier et al. state, theory often indicates a preference between these models; however the statistical criteria by which these are delineated offer much less clarity. In many ways the procedure that they are discussing is reminiscent of the instrument validation process. In both cases it is not a search for the truth. von Davier et al. provided 6 types of evidence that might be considered when choosing a statistical model. The first of these is substantive theory. This must always be one driver of model selection, but researchers in an exploratory mode may have little theory on which to rely. So typically one turns to log-likelihood-related criteria in an attempt to choose an appropriate model; in this case von Davier et al. used information criteria, log-penalty, and the average reduction in log-penalty. Classification accuracy, specifically average classification error and the reduction in classification error, can also be examined as per the recommendations by Vermunt (2010). The fourth source of evidence von Davier et al. provided were class profiles across identified subpopulations, showing how simple pictorial representations can provide useful information in the model selection process. External information in the form of literacy scores was also utilized to help validate the choice of the number of classes. Finally, von Davier et al. discussed the idea of usefulness as it relates to the choice of model. This is an especially important consideration when large data sets are examined, because models with large numbers of classes can be estimated. These types of evidence are excellent as a starting point in the model selection process. There are other sources of evidence that can provide the research with complementary information to aid in choosing a model. In this commentary, the author highlights several additional possibilities. Researchers would be well served to appreciate that there are numerous models that might be useful and that sometimes these models will be based on different assumptions regarding the nature of the latent construct. In the words of von Davier et al., "the reality of data analysis is that if there are many competing models, there will be multiple models that fit the data more or less with the same level of (in-) accuracy" (p. 204). (ERIC). |
Anmerkungen | Psychology Press. Available from: Taylor & Francis, Ltd. 325 Chestnut Street Suite 800, Philadelphia, PA 19106. Tel: 800-354-1420; Fax: 215-625-2940; Web site: http://www.tandf.co.uk/journals |
Erfasst von | ERIC (Education Resources Information Center), Washington, DC |
Update | 2017/4/10 |