Hello there,
I have been trying to reproduce Cox & Oakes (1984) estimates for the Stanford heart transplant data (included in the book), maximising the partial likelihood for only one time-varying covariate
Z(t) = 1 if the patient has not received a heart by time t and 0 otherwise. The estimate that Cox gives for \beta_1 is -0.258 with a maximum value for the partial likelihood of -555.07.
I implemented a maximising routine in fortran an obtained an estimate value for the same parameter of -0.4308 with a maximum value for the partial likelihood of -806.80.
Has any of you had the same problem?.
Luis.
|