Search code examples
stanrstan

recompiling to avoid crashing R session

How to avoid recompiling?

My stan() is recompiling to avoid crashing R session.

To validate my model, I want to replicate various models for many data from known distributions. However rstan::stan() always recompile my stan file, so it took a very long time. How to avoid recompiling.

Edit for comment-------------

Memorandum for conversion from rstan::stan() to rstan::sampling().

In rstan::stan, its variable name is model_name

 m <- "foo.stan"
 rstan::stan( model_name =m, data =...)

On the other hand, in rstan::sampling its variable name is object

   m <- "foo.stan"
   m <- stan_model(m)
   rstan::sampling( object = m , data = ...)

Please be careful !! The variable name is different !!

Solution

  • Use the function stan_model to compile, then use the compiled model output in the sampling function. For example:

    m <- stan_model('foo.stan')
fit <- sampling(m, data = ...)

    instead of

    fit <- stan('foo.stan', data = ...)