ERIC Number: ED159218
Record Type: RIE
Publication Date: N/A
Reference Count: 0
A Stochastic Decision Model for Experiments in Discrimination Learning.
Egelston, Richard L.
Decisions to terminate training for subjects involved with discrete trial experiments in discrimination learning should utilize a probability criterion rather than a deterministic criterion. Furthermore, these decisions should be based upon the number of correct and error responses made by the subject, with the decision made to terminate training or continue training following each correct response of the subject. The decision model presented in this paper is based upon the binomial and negative binomial probability distributions. The model is particularly useful when subsequent discrimination training is performed, or when between-group comparisons are made. Since terminal runs of varying lengths will be found with different subjects for the same decision level, it becomes possible to experimentally control for the effects of overtraining by holding decision levels constant within groups of subjects. Application of the model is equivalent to testing the hypothesis that a terminal run of correct responses of length x could have occurred on the basis of chance, conditional upon the respective total numbers of prior correct and error responses. Eighty-three tables for four levels of significance are provided. Since the experimenter may wish to use decision levels different from those provided, a computer program is provided for producing tables to different specifications. (Author/CTM)
Publication Type: Reports - Research
Education Level: N/A
Authoring Institution: N/A
Note: Parts may not reproduce clearly; Best copy available