0

I am currently trying to implement the Prentice, Williams and Peterson extension of the Cox model in R on survival data but I want a parametric model and can not get it done. Is there maybe a package I am unaware of or has anyone else created a parametric PWP model before? I would greatly appreciate any help with this.

Thanks J

Jaco Schoeman
  • 1
  • Questions asking us to recommend a library or tool are off-topic for StackOverflow. Besides that this question is also overly broad. Please take a [tour]. – Sumurai8 Aug 29 '14 at 09:56

1 Answers1

1

I don't see why you would use a parametric model.

This means specifying parametric forms for baseline hazards (and in PWP you have one for every event).

Regardless, you can use maximum likelihood for this, but you will probably have to program it yourself.

If you denote the hazard for the j-th event by lambda[0] (j), then the contribution of an individual i would be:

enter image description here

Just choose your parametric form for the lambdas, take the log, and plug into optim().

dardisco
  • 5,086
  • 2
  • 39
  • 54
Theodor
  • 986
  • 3
  • 7
  • 23