As with the two-response experimental data, it is of interest to see whether we can discriminate among these models with real experimental data. Table 15.5 gives the RMSD for each model for each of the six subjects in the QG task, a fit of a mean subject, and the mean of the subject fits. The FLMP and theoretically equivalent (BPM, GMM-E, EMM-CB, TSD) models all had fits of about RMSD = 0.0426. Figure 15.29 shows the fit of the FLMP for the typical subject shown in Figure 15.19. Of the remaining models, none differed significantly from the FLMP in goodness of fit, except the nonthreshold FCM (RMSD = 0.2364), F(1, 5) = 371.1, p < 0.001, and the IAC-RGR5 (RMSD = 0.0804), F(1, 5) = 111.5, p < 0.001. For the IAC-RGRK model, the additional k parameter averaged 9.2 and for the IAC-BOW model the standard deviation of the input noise averaged 0.133.
Several models of categorization were developed and analyzed within the context of prototypical pattern recognition tasks. These tasks involve the independent manipulation of two sources of information. The subject categorizes the stimulus event by choosing among two- or four-response alternatives. This task has been
Questia, a part of Gale, Cengage Learning. www.questia.com
Publication information: Book title: Multidimensional Models of Perception and Cognition. Contributors: F. Gregory Ashby - Editor. Publisher: Lawrence Erlbaum Associates. Place of publication: Hillsdale, NJ. Publication year: 1992. Page number: 442.
This material is protected by copyright and, with the exception of fair use, may not be further copied, distributed or transmitted in any form or by any means.