Maximum-Likelihood Sequential Procedure for Estimation of Psychometric Functions

Publisher Website
Google Scholar
Abstract
This paper describes a computer‐oriented sequential procedure for the estimation of parameters of a psychometric function. In common with other sequential procedures (BUDTIF, PEST, UDTR), the goal is to concentrate observations in the region of interest. The experimenter assumes a parametric form of the psychometric function (cumulative normal from 0 to 1, for example, but any other desired form could be used). After each response, values of the parameters are computed that maximize the probability of the set of responses that have been obtained, given the set of stimuli that have been presented. These values are then used to determine the placement of the next stimulus. A compromise between fidelity of tracking a time‐varying threshold and accuracy can be effected by appropriate choice of the weighting function for previous observations. The procedure has been simulated on a GE‐645 computer and applied in the laboratory with a DDP‐516 computer.

This publication has 0 references indexed in Scilit: