Note that each of these models is rather well able to predict subjects’ fear of movement-related pain ratings,
as they explain 77% of the variance in the observed fear of movement-related pain reports. Including a
random subject effect in the model to account for different average fear of movement-related pain ratings
across subjects is clearly important, because for each model, a considerable part of the variability in the fear
of movement-related pain ratings is due to differences between subjects. In particular, for each model,
s2 q=ðs2 q1s2 ε Þ ¼ :53, which means that 53% of the variability in the ratings is due to differences among subjects.
Furthermore, when omitting the random subject effect, the variance explained in the fear of movementrelated
pain rating drops from 77% to 48% for each of the models.