For each subject, the data set contains ten successive measurements (for days 0 through 9). The subject ID is used as a pseudo-covariate for model based partitioning here: The subject IDs are indicated in the tree structure in Figure 9, where the leftmost node, e.g., includes only the measurements of subject 309, while the rightmost node includes the measurements of subjects 308 and 350. The model of interest in each final node relates the number of days of sleep deprivation (0 through 9, on the abscissa) to the reaction time (on the ordinate).
