Model object not written


#1

Please also provide the following information in addition to your question:

  • Operating System: Windows Server 2016
  • rstanarm Version: 2.18.1

This Windows server has 64 GB RAM with 104 (of 198) free GB storage.

Using a terminal session within RStudio to run a stan_lmer model.

The data frame has ~ 1.1 M observations.

R code is:


t_prior <- student_t(df = 7,
location = 0,
scale = 2.5)

StandardizedOME.stan_lmer <- stan_lmer(
StandardizedOME ~ I(Age - 55) + I(AnesthesiaDuration - 140) + AnesthesiaTechniqueBlock +
AnesthesiaTechniqueGeneral + AnesthesiaTechniqueNeuraxial + ASAClass +
EmergencyStatusYN + Race + Sex + REMI + NonOpioidAnalgesicsCount +
I(DateOfService - 2015) + CPTBucket + (1 | MPOGInstitutionID),
data = AIM1Small.df, subset = StandardizedOME != 0,
prior = t_prior,
prior_intercept = t_prior,
chains = 4,
cores = 4
)

saveRDS(StandardizedOME.stan_lmer, ‘RDS/StandardizedOME.stan_lmer.Rds’)


The model started and ran about 10 days with appropriate terminal messages: 4 parallel processes started for the 4 chains, gradient evaluation, iteration warmup, iteration sampling, etc.

At the 10th day Chain 3 finished and reported elapsed times.

No messages about the other 3 chains was reported.

This was followed by an error message:

saveRDS(StandardizedOME.stan_lmer, ‘RDS/StandardizedOME.stan_lmer.Rds’) :
object ‘StandardizedOME.stan_lmer’ not found

The data frame and the t_prior object are present in the terminal session.

There is no model object.

I don’t understand this failure to finish all chains and write a model object.

Is my code incorrect?

Have I overwhelmed the storage space?

Anything else?

Comments appreciated.

Nathan