Prev: using inline and ODE45
Next: Equivalent of fftshift for Radon Transform
From: leo nidas on 20 Aug 2008 16:48

Hi there,

I just learned to use coxphfit and studying a bit about the
cox model. I cannot understand why in the cox model there
is no intercept? And why couldn't we model the hazard ratio
as linear function of the covariates?Any help from anyone?

Thanks in advance!
From: Tom Lane on 21 Aug 2008 14:17
> I just learned to use coxphfit and studying a bit about the
> cox model. I cannot understand why in the cox model there
> is no intercept? And why couldn't we model the hazard ratio
> as linear function of the covariates?Any help from anyone?

Leo, if I understand your questions correctly, the answers are:

1. If you included a constant, you'd adjust all hazards by the same factor.
This factor can be absorbed into the baseline hazard, so that kind of takes
the place of the constant.

2. You could use a linear function in principle to multiply the baseline
hazard, though the coxphfit function supports only a factor of the form
exp(X*b). The Cox and Oakes survival analysis book points out that this is
most common for good reasons, but others are possible including the linear
one.

-- Tom


 | 
Pages: 1
Prev: using inline and ODE45
Next: Equivalent of fftshift for Radon Transform