ABSTRACT
FIGURE 5.4: Empirical Bayes estimates of the subject-specic slopes 2+bi1 for the PBC dataset.
instantaneous risk for death. As it can been in Figure 5.4, the very large value for the slopes' coecients is explained by their rather small variability. Thus, it will not be meaningful to compute a hazard ratio for one unit increase in the slopes because an increase of such magnitude is not realistic. A more useful estimate for the subject-specic slopes can be obtained by dividing them with their estimated standard deviation from model jointFit6.pbc that equals 0.18. The updated joint model t is produced with the syntax
The results suggest that for patients receiving the same treatment and having the same hepatomegaly status, an increase of one standard deviation in the slope for log serum bilirubin is associated with a 5.82-fold increase (95% CI: 3.91; 8.67) in the hazard for death or transplantation.
