[R] proportional hazard with parametric baseline function: can it be estimated in R

[Thomas Lumley]

On Thu, 20 Dec 2001, Carl Mason --- Director Demography Lab wrote:

>
> Greetings --
>
> I would like to estimate a proportional hazard model with a weibull or
> lognormal baseline.  I have looked at both the coxph() and survreg()
> functions and neither appear (to me ) to do it.  Am I missing something in
> the docs or is there another terrific package out there that will do this.
>

survreg() will fit accelerated failure models with Weibull or lognormal
baselines. For the Weibull case these are equivalent to proportional
hazards models.  To convert the accelerated failure coefficients into
proportional hazards coefficients one just divides by -estimated scale,
which is given in the standard output.

Lognormal proportional hazards models are an unusual choice as the
lognormal family is not a proportional hazards family. If the distribution
is lognormal for all covariates=0 it will not be lognormal at any other
covariate value. This means that eg centering the covariates changes the
model.

Unless censoring is strongly related to your covariates or you want to use
the model for extrapolation to longer followup than you observed there is
no real advantage to a parametric proportional hazards model, so you might
as well use coxph().

	-thomas


-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._