Möbus, Claus and Wallasch, Roland (1977) Zur Erfassung von Hirnschädigungen bei Kindern : nichtlineare Entscheidungsregeln auf der Basis von Veränderungsmessungen und des Jackknife. Diagnostica, 23 (3). pp. 227-251. ISSN 0012-1924

- Published Version

Volltext (4Mb)
Official URL: http://www.hogrefe.de/zeitschriften/diagnostica/in...


In the prediction of minimal brain damage we try to combine the advantages of the multivariate and univariate approaches. Though we observe high hit rates using a battery of tests, we have to realize the instability of the results when we start a crossvalidation. The univariate approach offers stability but shows a lack in precision. This is absolutely undesirable for making individual-centered prognoses. We try to avoid these disadvantages with a nonlinear approach containing measures of change. A prognosis is made only in the case of the simultaneous appearence of three indicators (conjunction of indicatora). These indicators (high error rate in the pretest, remarkable change from pre- to posttest, high error rate in the posttest) are combined in a multiplicative way to a single predictor which can be put into a regression or discriminant equation. The results are very promising. We observe a conditional hit probability P[diagnosis = braindamaged | test = braindamaged] >.80 which can be crossvalidated by the leave-out-one (jackknife) technique of Mosteller & Tukey and Lachenbruch. Whereas in classical diagnostics researchers are only trying to maximize hit rates we want to increase the trustworthiness of the prognosis too. The trustworthiness increases with the critical distance of the prognosis to the cutoff. It can be shown that rules of prognosis with similar hit rates show large differences in their trustworthiness.

Item Type: Article
Uncontrolled Keywords: nonlinear decision rules, measurement of change, crossvalidation with leave-out one technique, jackknife technique, statistical prognosis of minimal brain damage, multiplicative combination of predictors, pre-postest change
Subjects: Generalities, computers, information > Computer science, internet
Philosophy and psychology > Psychology
Divisions: School of Computing Science, Business Administration, Economics and Law > Department of Computing Science
Date Deposited: 19 Jan 2016 10:59
Last Modified: 19 Jan 2016 10:59
URI: https://oops.uni-oldenburg.de/id/eprint/2575
URN: urn:nbn:de:gbv:715-oops-26566

Actions (login required)

View Item View Item

Document Downloads

More statistics for this item...