BAYES DECISION PROCEDURES FOR STIMULUS SAMPLING MODELS: I. NONSEQUENTIAL EXPERIMENTATION.
SYSTEM DEVELOPMENT CORP SANTA MONICA CALIF
Pagination or Media Count:
In stimulus sampling models of learning, the probability distribution defined on the sequence of conditioning functions which are used in these models may be regarded as a distribution over parameters. Consequently, this probability distribution is interpreted as an a priori distribution and the appropriateness of Bayes decision procedures for solving statistical decision problems involving these models is shown. Using beta distributions as a tractable family of prior distributions over the parameters of the single element model, Bayes solutions are illustrated to 1 the learning criterion problem, 2 parameter estimation problems, and 3 the optimal design of a learning experiment. Author