File:Developing-Bayesian-adaptive-methods-for-estimating-sensitivity-thresholds-(d′)-in-Yes-No-and-Video2.ogv

English: The qYNC method applied in cued detection. The upper panels present the true and estimated psychometric functions (left), the trial sequence (middle), and calculation of the pre-trial stimulus selection algorithm (right). The lower panels present sensitivity and decision parameter estimates, via the 2-d marginal pdfs defined over (1) sensitivity parameters (threshold and steepness), and (2) decision criteria: the strict criterion, and the z-score difference between the strict and lax criteria. The empirical psychometric functions generated by the liberal and conservative response states are presented for the true observer (blue) and for qYNC estimates (red). In the simulated trial sequence, the contrast of the selected stimulus and the cued response state are presented as a function of trial number. For each trial, the dot's position on the abscissa represents the trial number and the position on the ordinate represents the selected contrast. The outline color signifies the cued response state (red-strict; green-lax) and the dot's filled color signifies the observer's response (blue-Yes; black-No). The shaded regions represent the 65% (dark red) and 95% (lighter red) confidence intervals for sensitivity threshold estimates, as a function of trial number. The peak of the information gain function for the conservative response state is shifted to higher signal contrasts, relative to that of the liberal state. In addition, at lower signal intensities, the stimulus selection function falls off less slowly for the liberal state. The trial sequence demonstrates the qYNC method systematically samples different stimulus regions for the liberal and conservative response states. As a result of this stimulus selection pattern, the quick YNC is able to sample the dynamic regions of both empirical psychometric functions, and thereby sample a broader range of detection behavior than is possible in simple detection.