ABSTRACT
Considering patients' prognoses serves to adjust for treatment assignments that are not balanced with respect to prognosis. Suppose that patient 1 is assigned to treatment I, patients 2 and 3 are assigned to treatment P, and that WI is twice as large as W2 = W3. Then if all three patients have been in the study for the same amount of time (tl = t2 = t3) and none of the three patients have died, patients 2 and 3 combined make the same contribution to the likelihood function of Ap that patient 1 alone makes to that of AI.
