Figure 4

Participant-level (left) and study-level (right) random effect estimates for accelerometer (top) and GPS (bottom). The dots are the mean random effect estimates and the bars are the 95% credible intervals for each participant or study. Credible intervals that include 0 are shaded in grey while those that exclude 0 are shaded in black. The values on the y-axis represent the deviation from the overall average rate of sensor non-collection. There is substantial participant-level variation in missingness, and fairly low study-level variation relative to the participant-level variation. In all panels, estimates have been ordered from lowest (i.e., least sensor non-collection) to highest (most sensor non-collection) median value. Note that the y-axes differ across the rows.