R help - Jul 2006 - use "factor" for categorical covariate in Cox PH model

Hi All,

 

I'm learning the R codes for Cox PH modeling. Could I ask you what the
function of "factor" in modeling? Thank you!

 

When dealing with the categorical covariates (for example 3 groups), it
will come out different results if we add the command "factor" in
front
of the categorical covariate or not: if we don't add "factor",
there is
only one hazard ratio; if we add "factor", there are two hazard
ratios.

 

So does the "factor" actually create some dummy variables for the
calculation?

 

 

Here is the code example: 

 

curve2=survfit(Surv(CR, statcr) ~ factor(X1),data=dat1)

where X1 have three categories.

 

Thanks!

 


	[[alternative HTML version deleted]]

Linda Lei <llei <at> bccrc.ca> writes:
> 
> When dealing with the categorical covariates (for example 3 groups), it
> will come out different results if we add the command "factor" in
front
> of the categorical covariate or not 
The catch is here: the covariate is not a "categorial", but a number.
In that
case a regression is computed, and the ONE ration printed out has the meaning of
a slope.
> if we don't add "factor", there is
> only one hazard ratio; if we add "factor", there are two hazard
ratios.
> 
> So does the "factor" actually create some dummy variables for the
> calculation?
Factor make the covariate a categorial one, and two contrasts are printed out,
relative to the first level of the covariate.

Dieter