Slitting the Decision Maker's Throat with Occam's Razor -- The Superiority of Random Linear Models to Real Judges, No. 13
| dc.contributor.author | Dawes, Robyn M. | |
| dc.date.accessioned | 2025-01-29T21:46:18Z | |
| dc.date.available | 2025-01-29T21:46:18Z | |
| dc.date.issued | 1972-12 | |
| dc.description | 21 pages | |
| dc.description.abstract | Two types of linear models have been found to be superior to human judges in predicting codable criterion variables from codable predictor variables: actuarial models (based on the regression of the criterion in the predictors), and bootstrapping models (based on the regression of the judges• predictions on the predictors). In both types of models, the predictors are coded (or recoded) so that each bears a monotonic relationship to the criterion; further, in both models, nonlinear relationships and nonmonotone interactions between variables are ignored. Perhaps these linear models are superior because the expected validity of~ linear model whose weights are in the appropriate direction is superior to that of human judges in this context. Present research reviews a variety of situations in which linear models whose weights are in the appropriate direction but otherwise chosen at random do better than do judges , even boots trapped judges. | |
| dc.identifier.uri | https://hdl.handle.net/1794/30364 | |
| dc.language.iso | en_US | |
| dc.publisher | Oregon Research Institute | |
| dc.rights | Creative Commons BY-NC-ND 4.0-US | |
| dc.subject | linear models, criterion variables, codable output, multivariate codable input, output values | |
| dc.title | Slitting the Decision Maker's Throat with Occam's Razor -- The Superiority of Random Linear Models to Real Judges, No. 13 | |
| dc.type | Other |