Fig. C.2.

Temporal characteristics of 10 HMM states inferred in the simulated datasets compared to those in the real datasets. In each simulated multivariate dataset, fractional occupancy (A), fractional count (B), and mean lifetime (C) were computed for each state, for the sets of data points corresponding to the data of each CH participant. State-to-state transition probabilities (D) were also computed for the data-point sets corresponding to the data of each CH participant. We then determined values of these parameters corresponding to the 1st, 25th, 75th, and 99th percentiles in each simulated dataset. Boxplots show distributions of these values (as described in Fig. 6) across 50 datasets simulated based on the all-participant real dataset (in black) and 50 datasets simulated based on the CH-participant real dataset (in red). Corresponding parameter values in the all-participant real dataset (black solid circles) and in the CH-participant real dataset (red triangles) are shown for comparison. Inserts enlarge selective boxplots. The mean lifetimes of the HMM states were shorter in the real than simulated data (p < 0.05, corrected), and the within-participant variance in the other three temporal parameters between the HMM states was higher in the real than simulated data (p < 0.05, corrected).