Categories and dimensions
István Hidegkuti PhD
Faculty of Psychology and Educational Sciences
Department of Psychology, K.U.Leuven, Belgium
Supervisor: Prof. dr Paul de Boeck
Project running from: 1 November 2002 – 1 November 2006
In psychology it is rather common to use categories, for example psychiatric categories. An important question is whether what is considered to be a category is really categorical and not just dimensional. This issue is approached in two ways: with a broadly used method that is called taxometrics, and with a new method that is called dimcat.
Taxometrics is a common term for several analysis techniques of which MAXCOV (MAXimum COVariance) was the first and most frequently used. In the case of a taxonic structure the covariance of two indicators along the sections of a third indicator reaches its maximum value in the section where the proportion of taxon and non-taxon group members is equal. If the structure is dimensional, the covariance of two indicators will be about equal along the sections of a third indicator.
Dimcat is based on item response modeling (IRT). In dimcat distinctions are made along two axes, which are as follows: the first differentiates between-category qualitative differences and between-category quantitative differences. The second differentiates within-category heterogeneity and within-category homogeneity.
In this project the extension of the dimcat approach is suggested in two ways. First, a gener–al–iz–ation of the framework to rating-scale data are incorporated and an extension to latent categories. Finally, we propose a comparison of the extended dimcat and taxometrics along two dimensions: heterogeneous vs. homogeneous categories and smooth vs. abrupt differ–ences between the categories.
Date of defence: 24 May 2007
Title of thesis: Categories and dimensions.