The MBR metric |
| |
| Affiliation: | 1. Petru Maior University, Nicolae Iorga Street 1, Tîrgu-Mureș 540088, Romania;2. Sapientia University, Corunca 1C, Tîrgu-Mureș 540485, Romania;1. Department of Environmental Engineering and Water Technology, UNESCO-IHE Institute for Water Education, Westvest 7, 2611AX, Delft, The Netherlands;2. JCI Industries, Inc., 1161 S.E. Hamblen Road Lee''s Summit, MO 64081, Kansas City, MO, USA;3. Department of Biotechnology, Delft University of Technology, Van der Maasweg 9, 2629 HZ, Delft, The Netherlands |
| |
| Abstract: | P. H. Schönemann's (1982, Bulletin of the Psychonomic Society, 19, 317–319; 1983, Journal of Mathematical Psychology, 27, 311–324) MBR metric is designed to account for distortion effects in distance estimates caused by upper bounds in the response scales. To test the MBR hypothesis, given data have to be transformed first into the half-open interval [0, 1). This is achieved by defining the greatest distance estimate as the upper bound, and then dividing all values by this bound plus some “small constant” e, which ensures that the interval is open. How e should be picked is left open. It is shown here that very small differences in the value chosen for e have massive effects on the fit of the MBR model. Moreover, if e → 0, then the fit will eventually get bad for all realistic data. |
| |
| Keywords: | |
| 本文献已被 ScienceDirect 等数据库收录！ |
|
