Fig. 5.

Computational model‐based analysis of the relationship between psychiatric factors (dimensions) and decision‐making processes. (a) Model comparison in the reward‐seeking task. Each bar denotes exceedance probability in Bayesian model selection⁶³ (higher values close to one indicate a better fit). RL, reinforcement learning. (b) Model comparison in the loss‐avoidance task. Same format as in (a). (c) Effects of RL parameters on the CIT factor in the reward‐seeking task (mean $\pm$ SEM, n = 939). *P < 0.05, FDR‐corrected by the number of tests (i.e. 3) in one‐tailed t‐tests (α: corrected P = 0.026 and uncorrected P = 0.009; β: corrected and uncorrected Ps = 0.166; γ: corrected P = 0.031 and uncorrected P = 0.021). CIT, compulsive behavior and intrusive thought. (d) Effects of RL parameters on the CIT factor in the loss‐avoidance task (mean  $\pm$  SEM, n = 961). Same format as in (c) (α: corrected P = 0.225 and uncorrected P = 0.078; β: corrected P = 0.225 and uncorrected P = 0.150; γ: corrected and uncorrected Ps = 0.300).