Search code examples
rbayesianstanrstan

Trouble loading Stan model output using readRDS

I am running a Bayesian latent variable model in Stan. Here is my code for the model:

# run the Stan model 
mod.dyn <- stan(file="DynamicLVMmin.stan",
                data=stan.data, seed=570175513, thin = 5,
                iter=10000)

The model runs with no issue. I then go to save the mod.dyn as follows:

# save stan fit 
saveRDS(mod.dyn, "dynamic_fit.rds")

Now I am reading dynamic_fit.rds. I want to assess model convergence using the Gelman-Rubin R-hat diagnostic. Here is the code:

readRDS("dynamic_fit.rds")



params <- c("innov", "c_n_4_v2exdfdshs", "c_n_5_v2exdfvths", "c_n_3_v2exdfpphs", "c_n_5_v2lginvstp", 
            "c_n_3_v2lgoppart", "c_n_2_v2lgfunds", "c_n_4_v2lgcomslo", "c_n_2_v2lgintbup", "c_n_2_v2lgintblo", 
            paste("beta[", 1:9, "]", sep=""))

params.all <- names(mod.dyn)

However, I receive the following error:

Error: object 'mod.dyn' not found
Execution halted

What am I missing here? Shouldn't mod.dyn have been saved when I used saveRDS(mod.dyn, "dynamic_fit.rds"). Any feedback would be appreciated.

Solution

  • readRDS saves a single object without a name. So you want to use

    mod.dyn <- readRDS("dynamic_fit.rds")

    to capture that value after you read it in. This is different than save()/load() which store one or more variables with their name+value in a serialized format. If you did

    save(mod.dyn, file="dynamic_fit.rdata")

    then

    load("dynamic_fit.rdata")

    the object mod.dyn would be loaded back into your workspace.