Adaptive Psychophysical Procedures: Clarification and a New Algorithm
by Tony Wirth
Abstract: The problem of finding the threshold of a psychometric function is of major interest to researchers studying perception. This function models the response of a subject to different kinds of stimuli or, more often, stimuli of varying intensities. Essentially a problem of estimating the location or shape of a function, the estimation
procedure can be adaptive, therefore presenting us with an experimental design task. There is much similarity to the problems of binary search with errors and sequential estimation.
In our paper, we clarify the measures of accuracy and performance of various existing methods. We also present a new algorithm for
estimating the threshold that is a modification of an
information-theory inspired method. Our algorithm appears to be robust to inaccuracies in the assumptions, outperforming existing methods even in these circumstances.
Joint work with Elena Kelareva BSc Hons 06, James Mewing BE BCS 08
and Andrew Turpin RMIT University.
For More Information: Dr Owen Jones O.D.Jones@ms.unimelb.edu.au