Figure 2
Model output and diagnostics for A_GLM1 and B_GLM1. (a,b) Predicted probabilities (y-axis) of the models grouped by condition (x-axis). The error bars show the 95% confidence interval of predicted probabilities for the fixed effects computed from the whole models (taking random effects into account). The significance levels indicated by the asterisks represent Bonferroni-corrected least square means comparisons. (a) Shows the predicted probabilities of placing a bet in Version A of the paradigm obtained from model A_GLM1. (b) Shows the predicted probabilities of guessing “win” in Version B of the paradigm as obtained from model B_GLM1. (c,d) Show the ROC-plots of the models as a measure of model performance. The curves plot the true positive rates of the model on the y-axis against the false positive rates on the x-axis as computed on the test set for different discrimination thresholds. A curve leaning towards the top left, away from the diagonal, indicates a good model performance and a curve leaning towards the bottom right would point towards a poorer model performance with more false positives and false negatives. The integral of the curve (AUC) is an additional measure of model performance. (c) Shows the ROC-plot for model A_GLM1 as obtained by testing the model on the test set for Version A of the paradigm. (d) Represents the ROC-curve for model B_GLM1 as tested on the test set for Version B of the paradigm. (e,f) Plot the estimates of random effects of the mixed models together with their 95% confidence intervals with all participants on the y-axis and the estimates on the x-axis. (e) Shows the random effects in model A_GLM1. (f) Plots the random effects in model B_GLM1.
