Figure 4. Comparing differential posterior predictive log-likelihoods (ΔPPL) between DDM-equivalent model and exact input model (ExaM), with standard accumulation.

The difference in posterior predictive log-likelihoods (PPL) for each condition and each trial, averaged over 24 participants. PPLs of the DDM-equivalent model were subtracted from PPLs of the ExaM. A positive value (green) indicates that the ExaM has higher PPL than the DDM-equivalent, and a negative value (yellow), the opposite. The vertical black bars are the standard error of mean. The vertical red dotted line indicates the number of trials which the ExaM was better, with the number of trials indicated below.