It's All About the Subject - Options to Improve Psychometric Procedure Performance

摘要

We investigate the effect of procedure-specific parameters on the performance of three common psychophysical procedures. Methods considered include transformed-staircases, the Ψ-method and the UML method, while performance is evaluated in terms of accuracy, efficiency, precision and robustness. Simple Yes/No- and three alternative forced choice response paradigms were considered. A Monte Carlo simulation was conducted for three different types of test persons and analyzed by analysis of variances. Results show a large effect of the test person on the performance, especially for staircase procedures. No parameter exhibited a relevant effect on accuracy for each analyzed methods, estimation precision can be increased with an increasing number of trials. Only for staircase procedures, efficiency can be influenced by the choice of the progression rule.