Figure 6.

Simulation using the Bayesian forecasting procedure to predict performance over time for subject “A,” under conditions of sparse data availability. A large portion of the original data set (see Figure 2, left column panels) was discarded here, so as to simulate that the available performance measurements occurred infrequently—at random, unequally spaced intervals. Other details are the same as for Figure 2, except that the first panel in that figure (awake 4 h) is not repeated because only one data point was available during the first 4 h of wakefulness in this simulation.