Fig. 4.

Summary measures of Bayesian model performance. As the first five iterations were used as a burn-in for a first estimate of the Bayesian model, they are not depicted here. (a–b) Performance measures were derived at each iteration for each run by updating the model with every new observation made, i.e. incrementally re-simulating the online scenario. (c–d) Performance measures were derived when updating the model derived from the first run with each new observation made in succeeding runs from the same subject. Results are shown for four subjects as only one run was available for subject sub_04. (a) Mean SD across the predicted values of all possible 361 audio–visual stimulus combinations for each run of every subject. Each color represents a single subject. (b) Mean Euclidean distance between predicted optimum (coordinate with maximum predicted value) from hypothesized optimum (i.e., [10 10]) across all subjects for each run. Each color represents a different run. Shaded areas represent the SEM. (c) Mean SD across predicted values of all possible stimuli combinations at each iteration for concatenated runs. (d) Euclidean distance of predicted optimum from hypothesized optimum at each iteration for concatenated runs. Each color represents a single subject. Note, that at the final update (at iteration 76), the results are the same as those depicted in Fig. 3a–b.