Efficient and unbiased modifications of the QUEST threshold method: theory, simulations, experimental evaluation and...

Area of Science:

• Psychophysics
• Computational Neuroscience
• Statistical Modeling

Background:

• The QUEST method provides an efficient Bayesian approach for measuring perceptual thresholds.
• Modifications to QUEST may improve precision and reduce bias in threshold estimation.

Purpose of the Study:

• To evaluate variations of the QUEST method, focusing on stimulus intensity selection and threshold criteria.
• To compare the precision and bias of different QUEST modifications, including ZEST and the Minimum Variance Method.

Main Methods:

• Computer simulations evaluated four QUEST variations, including using the mean (ZEST), median, and mode of the probability density function (p.d.f.) for stimulus selection.
• The Minimum Variance Method was introduced, selecting stimulus intensity to minimize expected variance.
• Exact enumeration was used for up to 20 trials in yes-no and two-alternative forced-choice (2AFC) experiments.

Main Results:

• ZEST (using the mean) demonstrated superior precision compared to using the median or mode.
• The Minimum Variance Method offered slightly better precision than ZEST.
• Optimizing the threshold criterion, rather than relying on the 'ideal sweat factor,' can significantly improve measurement efficiency.

Conclusions:

• ZEST with an optimized threshold criterion or the Minimum Variance Method are recommended for precise threshold estimation.
• The study distinguishes between measurement bias and interpretation bias, noting that current methods have measurement bias if assumptions hold.
• The relative merits of yes-no and 2AFC techniques were also compared.