Determination of optimal data placement for psychometric function estimation: A computer simulation

Psychometric functions are used to relate the responses of a subject to phy sical stimuli in a variety of psychophysical tasks. However, it is time con suming to obtain data to determine a psychometric function if many stimulus levels and many trials are required. A computer simulation was conducted t o determine the minimum number of data points needed-for such a determinati on. The computer simulation also determined the optimal placements of the s timuli and the number of trials per datum point for psychometric function d eterminations. Results indicate that a 2-point sampling method with 30-50 t rials per point at optimal locations can produce a psychometric function wi th accurate spread and threshold estimates in a yes-no paradigm. However, t he 4-point sampling method yields statistically smaller variances of the es timates. For the 2-alternative forced-choice paradigm, at least 120 trials per point are needed for the 2-point sampling method's estimated parameters to differ from the known parameter values by less than 5%. The simulation results suggest that 3-alternative or 4-alternative forced-choice is prefer able to 2-alternative. Furthermore, when a criterion-free paradigm is not r equired, the yes-no paradigm is a better procedure than m-alternative force d-choice for obtaining the corresponding psychometric function because of s maller standard deviation of the estimates and smaller number of trials/poi nt required. (C) 1999 Acoustical Society of America. [S0001 -4966(99)02909- 4].