Hi,
I have the IMSL Fortran Numerical Library Version 6.0 installed on my computer and I am using the Intel Visual Fortran Compiler 10.1.011 for Windows.
I am having trouble using the PHGLM subroutine to estimate a proportional hazards model. Specifically, could you tell me how to estimate a proportional hazards model when I have multiple records per subject?
For example, I am looking at when a person retires, and I have multiple observations of the individual through his retirement. The data for a given individual (pid) look like:
pid ageretire x1
1 55 0 14
1 56 0 18
1 57 0 0
1 58 1 25
With multiple individuals, say 4, the data looks like:
pid age retire x1
1 55 0 14
1 56 0 18
1 57 0 0
1 58 1 25
2 55 0 14
2 56 0 18
2 57 1 26
3 55 0 25
3 56 0 14
3 57 0 18
3 58 1 30
4 55 0 12
4 56 0 14
4 57 1 18
In this case, I would like to estimate a proportional hazard model and obtain the coefficient on x1 to see how x1 predicts retirement.
Can you tell me how I can specify that: (1)the analysis time is age; (2)the failure event is retire == 1; and (3) individuals are indentified by pid?
Can I do this with the PHGLM command or is there a different command I should use or is there some way I should transform the data?
I appreciate any help. Thanks!
