Validierung von Fähigkeitstests auf Grundlage stochastischer Testmodelle

Abstract

Stochastic testmodels open up completely new possibilities of test construction.In the following thesis, the three psychometric tests Benton, Raven and Cattell were subjected separately and as a test battery to detailed classical and probabilistic test-theoretical examinations. The comparison of these two methods regarding the forecast validity of clinical criteria had become necessary for the cross validation of test-diagnostically remarkable results from Kindermann (2001).In this previous work, two latent classes were identified using a new form of the Benton test with the mixed Rasch-model. The surprising result was that 93% of tested patients in a mixed random sample fell into only one of the two identified classes. This result may suggest that two qualitatively different and diagnostically relevant solution strategies within the Benton test are identifiable only with stochastic test models of modern theory.For the current thesis, a completely new set of data was collected. The total sample (N=615) was comprised, as in Kindermann (2001), of clinically symptomatic and asymptomatic persons with psychiatric and neurological diagnoses.Utilizing mixed distribution models in the cross validation, it was possible to identify qualitative information within the data of each individual test, independent of the test performance.With regard to the previously almost unnoticeable or nearly non-existent relevance of modern test theory in practical test-diagnostic applications, the following conclusions can be made from the considerations in this thesis:The classical approach for measuring psychological abilities in the present interpretation proves to be inconsistent, despite the fact that this theory appears to be test-theoretically correct.Although quantitative test diagnostics, according to the theory of Rasch homogeneity technically measure theoretical given abilities in a person, they do not account for the subject role of each individual within the test, who is characterised by his or her freedom of thought, mental aliveness and uniqueness.Suggestions and related factors will be discussed in the following thesis.