Returning both group and individual level transformed parameters

Modeling
1

Hello, I am running a hierarchical model that transforms both individual- and group-level parameters, and it is supposed to return (1) an array of parameters fitted on the individual level, and (2) a mean and standard deviation term for each parameter calculated from the group level parameters. However, currently the model only returns the fitted parameters from the individual level. I am wondering why I cannot access the mean and standard deviation variables (ending in _mu and _sd) from the stanfit object, even though they are declared in the parameters {} block?

Here is my model:

data {
  int ntr;          // number of total data points (N)
  int nsub;         // total number of participants
  int ParticipantID[ntr];  // which subject each data point belongs to
  int nResponseTypes; // K
  real var1[ntr];
  real var2[ntr];
  real proximity[ntr]; //used to for both var3 and var4
  int actionTaken[ntr];  //y 
}

parameters {
  // Group level
  real<lower=0> invTemp_mu; 
  real<lower=0> invTemp_sd;
  real<lower=0> b_var1_mu;
  real<lower=0> b_var1_sd;
  real<lower=0> b_var2_mu;
  real<lower=0> b_var2_sd;
  real<lower=0> b_var3_sd;
  real<lower=0> b_var3_mu;

  // Individual participants
  real<lower=0> invTemp_pr[nsub];
  real b_var1_pr[nsub];
  real b_var2_pr[nsub];
  real b_var3_pr[nsub];
}

transformed parameters {
  // Group level transformed variables
  real b_var1_pr2[nsub];
  real b_var2_pr2[nsub];
  real b_var3_pr2[nsub];
  real<lower=0> k_mu;
  real<lower=0> b_var1_mu_divk;
  real<lower=0> b_var2_mu_divk;
  real<lower=0> b_var3_mu_divk;
  real<lower=0> b_var4_mu_divk;

  // Individual level transformed variables
  real k[nsub]; // sum of all absolute regression weights
  real b_var1[nsub];
  real b_var2[nsub];
  real b_var3[nsub];
  real b_var4[nsub];
  real invTemp[nsub];

  // First transform the priors to the group level
  for (is in 1:nsub) {
    invTemp[is] = exp(invTemp_pr[is] * invTemp_sd + log(invTemp_mu));  // what it does invTemp[is] ~normal(invTemp_mu,invTemp_sd)
    b_var1_pr2[is] = b_var1_pr[is] * b_var1_sd + b_var1_mu; 
    b_var2_pr2[is] = b_var2_pr[is] * b_var2_sd + b_var2_mu;
    b_var3_pr2[is] = b_var3_pr[is] * b_var3_sd + b_var3_mu;
  }

  // Then transform to the individual level using the group level transformations
  for (is in 1:nsub) {
    k[is] = fabs(1) + fabs(b_var1_pr2[is])+fabs(b_var2_pr2[is])+fabs(b_var3_pr2[is]);
    b_var4[is] = 1/k[is];
    b_var1[is] = b_var1_pr2[is]/k[is];
    b_var2[is] = b_var2_pr2[is]/k[is];
    b_var3[is] = b_var3_pr2[is]/k[is];
  }

  k_mu = fabs(1) + fabs(b_var1_mu) + fabs(b_var2_mu) + fabs(b_var3_mu);
  b_var4_mu_divk = 1/k_mu;
  b_var1_mu_divk = b_var1_mu/k_mu;
  b_var2_mu_divk = b_var2_mu/k_mu;
  b_var3_mu_divk = b_var3_mu/k_mu;
}

model{
  matrix[ntr,nResponseTypes] invTxUtil;

  // Priors - these become the priors for the group level
  invTemp_mu ~ cauchy(10,2); 
  b_var1_mu ~ cauchy(0,1);
  b_var2_mu ~ cauchy(0,1);
  b_var3_mu ~ cauchy(0,1);
  invTemp_sd ~ cauchy(0,3);
  b_var1_sd ~ cauchy(0,1);
  b_var2_sd ~ cauchy(0,1);
  b_var3_sd ~ cauchy(0,1);

  // individual subjects
  invTemp_pr             ~ normal(0,1);
  b_var1_pr ~ normal(0,1);
  b_var2_pr ~ normal(0,1);
  b_var3_pr ~ normal(0,1);

  // Iterate through each trial
  for (itr in 1:ntr) {
    // Utility for action1
    invTxUtil[itr,1] = invTemp[ParticipantID[itr]]*(b_var1[ParticipantID[itr]]*var1[itr]);
    // Utility for action2
    invTxUtil[itr,2] = invTemp[ParticipantID[itr]]*(b_var4[ParticipantID[itr]]*proximity[itr]);
    // Utility for action3
    invTxUtil[itr,3] = invTemp[ParticipantID[itr]]*(b_var2[ParticipantID[itr]]*var2[itr] + b_var3[ParticipantID[itr]]*proximity[itr]);
    actionTaken[itr] ~ categorical_logit(to_vector(invTxUtil[itr])); // link utilities to actual choices
  }
}

I appreciate any help, thank you.

2

Which interface are you using? If you’re using R, can you share your code? In particular, what is the output of summary(stanfit)[1]. You can also share your output file.