Dimension error in hierarchical reinforcement learning model

Modeling
1

Hi everyone,

I am new to stan and still trying to figure out the language. I followed instructions online for an exercise, but it seems that stan has updated and some of the language logic has changed a bit.

Please see the replicable code and the error below, and give me some advice. Thank you very much.

d=data.frame(
subjID=c(rep(1:10,each=100)),
ntrials=rep(1:100,10),
choice=sample(c(1,2),size=1000,replace = T),
outcome=sample(c(1,-1),size=1000,replace = T)
)


library(cmdstanr)

stan_file=write_stan_file('

data{
int<lower=1,upper=10>nSubjects; 
int<lower=1,upper=100> nTrials; 
array[nSubjects,nTrials] int<lower=1,upper=2> choice; array[nSubjects,nTrials] int<lower=-1,upper=1> outcome;
}

parameters{
  real<lower=0,upper=1> alpha_mu; real<lower=0> alpha_sd;     
  real<lower=0,upper=20> tau_mu;  real<lower=0> tau_sd;      
  array[nSubjects] real<lower=0,upper=1> alpha; 
array[nSubjects] real<lower=0,upper=20> tau; 
          }
         
model{
      alpha~normal(alpha_mu,alpha_sd); tau~normal(tau_mu,tau_sd); 
      alpha_sd~cauchy(0,1);tau_sd~cauchy(0,20);
      
      for (s in 1:nSubjects){

        vector[2] v=rep_vector(0,2); 
                                       
        for(t in 1:nTrials){
                           
              choice[s,t]~categorical_logit(tau[s]*v); 
              v[choice[s,t]]+=alpha[s]*(outcome[s,t]-v[choice[s,t]]); 
                            
                            }
                            }
      }
'
)

stan_d=list(nSubjects=10,nTrials=100,choice=d$choice,outcome=d$outcome)
stan_m=cmdstan_model(stan_file)
fit=stan_m$sample(data=stan_d)



Running MCMC with 4 sequential chains...

Chain 1 Exception: mismatch in number dimensions declared and found in context; processing stage=data initialization; variable name=choice; dims declared=(10,100); dims found=(1000) (in 'C:/Users/HUAWEI/AppData/Local/Temp/Rtmp8c0HJt/model-5490189352e5.stan', line 6, column 0 to column 53)
Warning: Chain 1 finished unexpectedly!

Chain 2 Exception: mismatch in number dimensions declared and found in context; processing stage=data initialization; variable name=choice; dims declared=(10,100); dims found=(1000) (in 'C:/Users/HUAWEI/AppData/Local/Temp/Rtmp8c0HJt/model-5490189352e5.stan', line 6, column 0 to column 53)
Warning: Chain 2 finished unexpectedly!

Chain 3 Exception: mismatch in number dimensions declared and found in context; processing stage=data initialization; variable name=choice; dims declared=(10,100); dims found=(1000) (in 'C:/Users/HUAWEI/AppData/Local/Temp/Rtmp8c0HJt/model-5490189352e5.stan', line 6, column 0 to column 53)
Warning: Chain 3 finished unexpectedly!

Chain 4 Exception: mismatch in number dimensions declared and found in context; processing stage=data initialization; variable name=choice; dims declared=(10,100); dims found=(1000) (in 'C:/Users/HUAWEI/AppData/Local/Temp/Rtmp8c0HJt/model-5490189352e5.stan', line 6, column 0 to column 53)
Warning: Chain 4 finished unexpectedly!

Warning: Use read_cmdstan_csv() to read the results of the failed chains.
Warning messages:
1: All chains finished unexpectedly! Use the $output(chain_id) method for more information.
 
2: No chains finished successfully. Unable to retrieve the fit.

I also tried the hbayesDM way for this model, because the code is much simpler. The error I got looks like this:

> library(hBayesDM)
Loading required package: Rcpp


This is hBayesDM version 1.2.1


Warning message:
package ‘hBayesDM’ was built under R version 4.5.1 

> fit=bandit2arm_delta(data=d,adapt_delta = 0.8)
Error in is.null(data) || is.na(data) : 
  'length = 4000' in coercion to 'logical(1)'
2

Hi,

There is a mistmach in the declaration of your data,

choice=sample(c(1,2),size=1000,replace = T),

is defined as a 1d_array, but latter you say that this is a 2d_array

array[nSubjects,nTrials] int<lower=1,upper=2> choice;

I guess you should modify, this

choice=d$choice

to give a 2d array to stan, and not a 1d array

1 Like