Example of discrete-time mHealth data and fitted PH-HMM. Probabilities of being in the active state, screen-on counts, mean acceleration magnitude, and hour-of-day random intercepts are plotted against time (hours). Regression models: ⁠, were fitted for individual i and MAP estimates were calculated using final E-step probabilities ⁠. Random intercepts capture the diurnal rhythm of active–rest cycles, with active states mainly occurring between the hours of 6 am–10 pm. Large values of correspond with a high magnitude in the cyclic diurnal effects. This figure appears in color in the electronic version of this paper, and any mention of color refers to that version.