🚀 go-pugleaf

>  For each child I know whether or not a seizure was observed and,
> if one was observed, approximately how long after the head trauma (within one
> hour, one to 12 hours, within 12 hours to 3 days, etc).  If no seizure
> was observed, I know how long the child was observed, after which we have
> no idea what happened.

> The Questions:

> 1.  Is the proportional hazards model reasonable?  If not, what should
    I use?

You don't give enough information to decide on this question. Firstly, time
must be measured exactly, not grouped. Secondly, it is not clear whether time
to seizure matters to you or you just want to know if the seizure was observed
and what might have caused it. Thirdly, it is not clear if there is a problem
of censored observations which are the main reason for using survival analysis
techniques. If all the children were observed for a certain period of time and
those without seizures dismissed at that point than there's no need for
survival analysis. You have a group with seizures and a group without them
and you can use logistic regression to analyze your data.

I hope this helps

Janez Stare
Medical Faculty
Ljubljana, Slovenia