Monte-Carlo Analysis of Two Logical Premises to Avoid Probit Algorithms for Determination of Sensory Thresholds by Psychophysics
Speech and Hearing Sciences
It is often necessary to apply various logical principles to extract the underlying model behavior from empirical data that are usually corrupt with a variety of perturbations. For example, psychophysical data of stimulus strength versus detection rate usually appear nonmonotonous when known to be essentially monotonous in principle. Often confusing and usually avoided, Probit algorithms are traditionally needed to iteratively calculate an optimal regression curve for such nonmonotonous data and then deduce the 50% detection threshold level. Two logical premises are presented that enable monotonicity even with nonmonotonous observed rates. Monte-Carlo simulation demonstrated that this strategy not only finds a monotonic relation between detection rate versus stimulus intensity, but also significantly improves accuracy of estimation of the 50% threshold level. In conclusion, two logical premises can help to obtain uniform monotonicity of the sigmoid shaped psychometric regression curve, even without using conventional Probit iterations, leading to faster and simplified estimation of the 50% threshold point with greater accuracy. Further research is needed to evaluate clinical impact of this concept for improving current procedures of threshold estimation without increasing the number of actual observations for various psychophysical assessment of sensory thresholds.
Proceedings of Meetings on Acoustics
(2017). Monte-Carlo Analysis of Two Logical Premises to Avoid Probit Algorithms for Determination of Sensory Thresholds by Psychophysics. Proceedings of Meetings on Acoustics, 26(1), 1-6.
Available at: https://aquila.usm.edu/fac_pubs/16465