Fig. 4: Discretized trial-by-trial model of criterion updating.
a Behavior of the model for a 10-trial sequence; see main text for definitions. b Effect of Δspeedn − 1 on trial n psychometric curves based on discretized history model. Horizontal shifts were set by µn. Blue squares denote points of subjective equality (PSE) and the black dashed line illustrates PSE computed with all trials merged. c Prediction improvement (Brier score decrease) using the discretized history model with respect to the no-history model, for each of the nine stimuli in trial n. Blue line is the median and shaded area the interquartile range of the cross-validations. d Comparison of the area under the ROC curve (AUROC) of the discretized history model vs. the no-history model for different cross-validation test sets. The upper histogram shows the AUROC difference computed by considering predictions for all stimuli in trial n. The lower histogram shows the AUROC difference computed by considering predictions for Δspeedn = −1 (the stimulus value for which the model afforded the largest benefit). Note the different abscissae scales. Each AUROC value derives from one round of cross-validation and the histograms plot the outcome of 100 rounds. Source data are provided as a Source Data file.
