Model | I: Null model | II: MLE model | III: Reliability-weighting | IV: Full model |
---|---|---|---|---|
Parameters (n) | 8 | 5 | 11 | 17 |
R2 (mean) | 0.656 | 0.658 | 0.687 | 0.722 |
Relative BIC (sum) | 0 | 59.662 | 1501.077 | 3202.034 |
Expected posterior p | 0.111 | 0.111 | 0.111 | 0.667 |
Exceedance p | 0.014 | 0.014 | 0.014 | 0.957 |
Protected exceedance p | 0.026 | 0.026 | 0.026 | 0.921 |
I: In the null model, neither PSEs nor slopes depended on visual reliability or modality-specific report. II: In the MLE model, audiovisual PSEs and slopes were predicted based on unisensory variances as described in Eqs. (1) and (3). III: In the reliability-weighted integration model, PSEs and slopes depended on visual reliability unconstrained by MLE predictions. IV: In the full model, PSEs and slopes depended on visual reliability unconstrained by MLE predictions and modality-specific report (MR). R2, coefficient of determination, corrected for the binary response option (Nagelkerke, 1991). Relative BIC, Bayesian information criterion (i.e., an approximation to the model evidence) at the group level, i.e., subject-specific BICs summed over all subjects (BIC = LL − 0.5M × ln(N), where LL = log likelihood, M = number of parameters, N = number of data points) of a model relative to the null model (note that a greater relative BIC indicates that a model provides a better explanation of our data). Expected posterior p, probability that a given model generated the data for a randomly selected subject; exceedance p, probability that a given model is more likely than any other model; protected exceedance p, probability that one model is more likely than any other model beyond chance.