0
$\begingroup$

thanks for helping!

I am testing a repeated measures dataset (multiple rows per participant) in a multilevel Cox regression with time-varying predictors. I'm trying to see if effects vary by assessor. My variables are:

  • ID: participant ID (time independent)
  • dynamic: repeated measure test score (time dependent)
  • assessor: person doing assessment, nesting variable (time independent)
  • intervention : binary, whether or not dynamic was followed by intervention (time dependent)
  • entry: beginning of prediction interval (time dependent)
  • exit: end of prediction interval (time dependent)
  • flag: if survival event occurred during interval (time dependent)

I want to eventually test if interactions between dynamic and intervention vary by assessor, but starting with univariate tests, I have two questions.

First: I keep seeing two separate ways of specifying that I want the slope (of "dynamic")to vary by assessor. What is the difference between

model<-coxme(Surv(entry,exit,flag)~dynamic+(1+dynamic|assessor),data=data)

and

model<-coxme(Surv(entry,exit,flag)~dynamic+(1|assessor/dynamic),data=data)?

Second: Does it make sense to test a random coefficient for a binary variable? Would that use different syntax? Would random effects be interpretable or would I go by chi-square of the models?

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.