0
$\begingroup$ 
fit.tdc <- 
coxph(Surv(t1,t2,status) ~ Hemoglobingroup + hemoglobin + sex + 
diabet + cluster(id), data)
fit.tdc

Can the hemoglobin and hemoglobin group be put in the same model? Or would it be silly? my hemoglobin groups are made up of classes like Low-Low, medium-medium, high- high belonging to hemoglobin

Improve this question
$\endgroup$
4
  • 3
    $\begingroup$ If hemoglobin group is just a discrete version of a continuous hemoglobin measurement then yes it would be silly in my opinion, since group is simply a lossy encoding of the continuous measure. The only situation I can think of that the discrete version might be better is in the case of non-linearity, but you would be better served checking this and using (eg) a spline to model non-linearities in my opinion. $\endgroup$
    – alan ocallaghan
    Commented Jan 17, 2021 at 11:42
  • $\begingroup$ then I guess if I create the model as below, if I don't look at the spline $\endgroup$
    – Cgdmm
    Commented Jan 17, 2021 at 11:55
  • $\begingroup$ fit.tdc<-coxph(Surv(t1,t2,status)~hemoglobin+sex+diabet+cluster(id),data) $\endgroup$
    – Cgdmm
    Commented Jan 17, 2021 at 11:56
  • $\begingroup$ Looks about right yeah. You should also look at model diagnostics, eg cox.zph $\endgroup$
    – alan ocallaghan
    Commented Jan 17, 2021 at 13:44

0

Reset to default

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.