R goes really slow implemented Rstan. Don't know what happened

Interfaces RStan
, ,
1

Hi everyone,

My Rstudio gets stuck once the Rstan is implemented, for example the “Check” button on the Rstan interface takes forever to run. The “rstan::” function takes forever too. Also, after I library(rstan), “::” function for all the packages get stuck. When I create a new project and avoid any steps related to rstan, everything goes fine. I have tried to update my R and Rstudio to the latest version but it didn’t help.

I believe the code is proper. paste it anyway:

data{
  int obs;                      // number of observation in the long vector
  int N;                        // number of individuals
  vector [N] size_t0;           // size of focal plants
  vector [N] growth;            // growth of focal plants, response
  int n_nb[N];
  vector [obs] size_vector;        // leng vector of neighbor sizes
  vector [obs] dist_vector;        // long vector of pairwise distances
  int pos[N];
}
transformed data{
}
parameters{
  real alpha;                       // Intercept  
  real beta;                        // Effect of target plant size on response  
  real sigma;                       // global SD  
  real<lower=0> a1;                 // size dependent effect of the neighbor    
  real a3;                          // net crowding effect  
  real<lower=0> a2;                 // spatial scale  
  }
transformed parameters{
  vector[N] kernel;  
  vector[N] mu;
  vector[obs] size_vec;
  vector[obs] dist_vec;
  for (i in 1:obs){
      dist_vec[i]=dist_vector[i]^2; 
      size_vec[i]=size_vector[i]^a1;
    }
  for(n in 1:N)
      kernel[n]=sum(exp(log(segment(size_vec, pos[n], n_nb[n]))-
                          (segment(dist_vec, pos[n], n_nb[n])*a2)));
  
  for(n in 1:N)
      mu[n]=alpha+size_t0[n]*beta+a3*kernel[n];
}
model{
  alpha~normal(0,5);   // uninformed priors centered at zero;
  beta~normal(0,5);
  a1~normal(0,5);
  a2~normal(0,5);
  a3~normal(0,5);
  sigma~exponential(1);
  
  growth ~ normal(mu,sigma);
}

Operating System: MAC Catalina 10.15.6
RStan Version: 2.21.2
Session info:
setting value
version R version 4.0.3 (2020-10-10)
os macOS Catalina 10.15.6
system x86_64, darwin17.0
ui RStudio
language (EN)
collate zh_CN.UTF-8
ctype zh_CN.UTF-8
tz Asia/Shanghai
date 2020-12-27

Thanks!

2 Likes
2

Hi, Yangkang and welcome to the community!

if it’s any comfort, using cmdstan I can compile this model. Try installing cmdstanr,

install.packages("cmdstanr", repos = c("https://mc-stan.org/r-packages/", getOption("repos")))

and then install cmdstan with,

install_cmdstan()

and then compile model with,

m0 <- cmdstan_model("foo.stan")
1 Like
3

Thank you torkar. I will try it. But still a pity that I can’t use the “Check” button to confirm my code before running it.

4

That could be an RStudio problem so better check with them?

5

Yes, I will use CmdStan before I figure it out. If I’m the only one to have this problem, must be something wrong with my RStudio. Thanks for your advise.

6

There’s another recent thread where a few of us have reported the same thing. I suspect it’s an rstudio bug but haven’t figured out how to make a reproducible example for a hug report.

3 Likes
7

I have a similar problem with you some months ago. The reason is that rstan will try to download a script from GitHub which is inaccessible in China. See code on this page stanc.R.

Try rstan_options ( javascript = FALSE)

4 Likes
8

Thanks! It’s exactly the solution. Wish I can remember to type this every time : )

9

It should be a permanent option so you only need to type once. At least I just use it once.