Multilevel mixed effects survival analysis

I’ve just updated stmixed, which fits multilevel parametric survival models. This is a big update, because stmixed is now a shell file for merlin…which does most things (see here).

merlin does not have a simple syntax, because of its generality, so by having convenience wrapper functions, I can make certain classes of models much more accessible. stmixed is the first of these.

The pre-print on stmixed has been updated to reflect the major changes:

Crowther MJ. Multilevel mixed effects survival analysis: Estimation, simulation and application.

A list of the new features is below, along with installation instructions.


New features and changes

  • any number of levels, with any number of random effects at each level, can now be specified
  • distribution(rcs) has been added, which uses restricted cubic splines to model the baseline function on the log hazard scale
  • distribution(user) has been added, which allows you to specify your own survival model by using merlin’s utility functions. See an example using fractional polynomials to model the baseline log hazard function.
  • delayed entry is now supported (but experimental!)
  • time-dependent effects are now supported with all models through the tvc() option (and others), allowing the use of restricted cubic splines of log time to model effects flexibly
  • bhazard() is supported with all models, to envoke a relative survival model
  • accelerated failure time models are no longer supported, as they are not currently available within merlin


To install the new version of stmixed type:

net install stmixed, from(

It will be up on SSC soon.

As stmixed now requires Stata 14.2 or higher, I’ve made available the previous version as stmixed12, for those using Stata 12.1 up to 14.2. You can install it from here using:

net install stmixed12, from(

Any bug finds, just let me know.


comments powered by Disqus