Performance Simulation of Unforced Choice Paradigms in Parametric Psychometric Procedures

This paper shows an implementation of the Psi andUML (Updated Maximum Likelihood) methods to incorporate unforced choice paradigms (nAUC) and simulation results for repeatability, efficiency and accuracy. Parametric methods like Psi and UML promise higher accuracy and efficiency compared to classic and non-parametric methods and support fixed sets of stimuli. Unforced choice paradigms have shown similar erformance as forced choice paradigms but are expected to create less confusion for test subjects for low stimuli intensities. An implementation of an unsure test person is presented. Psi and UML methods are compared to the Unforced Weighted Up-Down method (UWUD) in two Monte Carlo simulations. Considered measures are Variation Coefficient for repeatability, Sweat Factor for efficiency and threshold bias for accuracy. Psi and UML seem suitable to be combined with nAUC paradigms. Variation coefficients are smaller than 0.08 (Psi) and 0.15 (UML) for YN, 2AFC and 3AFC paradigms. UML-based procedures show a bias less than 2 %, while Psi-based procedures exhibit paradigm-dependent bias up to 10 %. Psi is at least twice as efficient as UML for 40 simulated trials. Because of large bias and poor repeatability, only the combination of UWUD with a 3AUC-paradigm shows results comparable to parametric procedures with nAUC paradigms