Brms error, after re-installing rstan on Catalina - cannot close 'message' sink connection

[mmichel]

Hi

After dealing with the recent Catalina upgrade on Mac and getting Rstan to work again, I’m getting a new error message that I haven’t seen before when using brms

“Error in close.connection(zz) : cannot close ‘message’ sink connection”

I’m wondering if anyone else has encountered this problem and if so, how they fixed the issue.

I did the standard googling to find an answer, but I have had no luck. I’ve also tried uninstalling and re-instaling R which didn’t solve the problem

I appreciate any help you could give me

Thanks for your time

[bgoodri]

The sink thing happens when there is a compiler error. Try calling brm with stan_model_args = list(verbose = TRUE).

[mmichel]

Thanks for your reply.

I tried adding that argument, but I get the same error.

Oddly, when I first fixed the Rstan Catalina issue and tried to run a model, it ran perfectly, but ever since I’ve been getting this error

[bgoodri]

Try calling make_standata(...) and make_stancode(...) to get the list of data and Stan program and then call stan(..., verbose = TRUE) using those.

[mmichel]

I called make_standata(…) which seemed to work, but for make_stancode(…), I got the same error

“Error in close.connection(zz) : cannot close ‘message’ sink connection”

[bgoodri]

I think brms is generating Stan code that does not parse somehow. Try first calling

debug(rstan::stanc)

and then when you call brm it should eventually give you a prompt. At that point, you can call

writeLines(model_code)

to see if you can figure out why the Stan code is messed up. Or post it here if you can’t.

[mmichel]

Thanks so much for your help with this, I’m not an experienced coder. I followed your instructions, and below is the output I received

// generated with brms 2.10.0
functions {
}
data {
int<lower=1> N; // number of observations
vector[N] Y; // response variable
int<lower=1> K; // number of population-level effects
matrix[N, K] X; // population-level design matrix
// data for group-level effects of ID 1
int<lower=1> N_1; // number of grouping levels
int<lower=1> M_1; // number of coefficients per level
int<lower=1> J_1[N]; // grouping indicator per observation
// group-level predictor values
vector[N] Z_1_1;
int prior_only; // should the likelihood be ignored?
}
transformed data {
int Kc = K - 1;
matrix[N, Kc] Xc; // centered version of X without an intercept
vector[Kc] means_X; // column means of X before centering
for (i in 2:K) {
means_X[i - 1] = mean(X[, i]);
Xc[, i - 1] = X[, i] - means_X[i - 1];
}
}
parameters {
vector[Kc] b; // population-level effects
// temporary intercept for centered predictors
real Intercept;
real<lower=0> sigma; // residual SD
vector<lower=0>[M_1] sd_1; // group-level standard deviations
// standardized group-level effects
vector[N_1] z_1[M_1];
}
transformed parameters {
// actual group-level effects
vector[N_1] r_1_1 = (sd_1[1] * (z_1[1]));
}
model {
// initialize linear predictor term
vector[N] mu = Intercept + Xc * b;
for (n in 1:N) {
// add more terms to the linear predictor
mu[n] += r_1_1[J_1[n]] * Z_1_1[n];
}
// priors including all constants
target += normal_lpdf(b | 0,10;
target += student_t_lpdf(Intercept | 3, 13, 10);
target += student_t_lpdf(sigma | 3, 0, 10)
- 1 * student_t_lccdf(0 | 3, 0, 10);
target += student_t_lpdf(sd_1 | 3, 0, 10)
- 1 * student_t_lccdf(0 | 3, 0, 10);
target += normal_lpdf(z_1[1] | 0, 1);
// likelihood including all constants
if (!prior_only) {
target += normal_lpdf(Y | mu, sigma);
}
}
generated quantities {
// actual population-level intercept
real b_Intercept = Intercept - dot_product(means_X, b);
}

[bgoodri]

@mmichel

target += normal_lpdf(b | 0,10;

This is missing its closing parenthesis. It should be target += normal_lpdf(b | 0, 10);. If you tell @paul.buerkner how you called brm(), he can fix it. Otherwise, you can just copy that to a .stan file, fix the closing parenthesis, and call stan yourself.

[mmichel]

Great, thanks for your help, I will give that a go.

[skb]

@mmichel Any news? Any fix for this error yet?

@bgoodri I get the same error:
Error in close.connection(zz) : cannot close ‘message’ sink connection

When I call debug(Stan::stand) and then writeLines(model_code) I get the following output. I believe that the parentheses are as they should be. Can you help?

// generated with brms 2.10.0
functions {
/* cumulative-logit log-PDF for a single response

• Args:
• y: response category
• mu: linear predictor
• thres: ordinal thresholds
• disc: discrimination parameter
• Returns:
• a scalar to be added to the log posterior
  */
  real cumulative_logit_lpmf(int y, real mu, vector thres, real disc) {
  int ncat = num_elements(thres) + 1;
  real p;
  if (y == 1) {
  p = inv_logit(disc * (thres[1] - mu));
  } else if (y == ncat) {
  p = 1 - inv_logit(disc * (thres[ncat - 1] - mu));
  } else {
  p = inv_logit(disc * (thres[y] - mu)) -
  inv_logit(disc * (thres[y - 1] - mu));
  }
  return log§;
  }
  }
  data {
  int<lower=1> N; // number of observations
  int<lower=2> ncat; // number of categories
  int Y[N]; // response variable
  int<lower=1> K; // number of population-level effects
  matrix[N, K] X; // population-level design matrix
  real<lower=0> disc; // discrimination parameters
  // data for group-level effects of ID 1
  int<lower=1> N_1; // number of grouping levels
  int<lower=1> M_1; // number of coefficients per level
  int<lower=1> J_1[N]; // grouping indicator per observation
  // group-level predictor values
  vector[N] Z_1_1;
  // data for group-level effects of ID 2
  int<lower=1> N_2; // number of grouping levels
  int<lower=1> M_2; // number of coefficients per level
  int<lower=1> J_2[N]; // grouping indicator per observation
  // group-level predictor values
  vector[N] Z_2_1;
  int prior_only; // should the likelihood be ignored?
  }
  transformed data {
  int Kc = K;
  matrix[N, Kc] Xc; // centered version of X
  vector[Kc] means_X; // column means of X before centering
  for (i in 1:K) {
  means_X[i] = mean(X[, i]);
  Xc[, i] = X[, i] - means_X[i];
  }
  }
  parameters {
  vector<lower=0>[Kc] b; // population-level effects
  // temporary thresholds for centered predictors
  ordered[ncat - 1] Intercept;
  vector<lower=0>[M_1] sd_1; // group-level standard deviations
  // standardized group-level effects
  vector[N_1] z_1[M_1];
  vector<lower=0>[M_2] sd_2; // group-level standard deviations
  // standardized group-level effects
  vector[N_2] z_2[M_2];
  }
  transformed parameters {
  // actual group-level effects
  vector[N_1] r_1_1 = (sd_1[1] * (z_1[1]));
  // actual group-level effects
  vector[N_2] r_2_1 = (sd_2[1] * (z_2[1]));
  }
  model {
  // initialize linear predictor term
  vector[N] mu = Xc * b;
  for (n in 1:N) {
  // add more terms to the linear predictor
  mu[n] += r_1_1[J_1[n]] * Z_1_1[n] + r_2_1[J_2[n]] * Z_2_1[n];
  }
  // priors including all constants
  target += normal_lpdf(b[1] | 0,.1)
  - 1 * normal_lccdf(0 | 0,.1);
  target += normal_lpdf(b[2] | 0,.1)
  - 1 * normal_lccdf(0 | 0,.1);
  target += normal_lpdf(b[3] | 0,.1)
  - 1 * normal_lccdf(0 | 0,.1);
  target += normal_lpdf(b[4] | 0,.1)
  - 1 * normal_lccdf(0 | 0,.1);
  target += normal_lpdf(b[5] | 0,.1)
  - 1 * normal_lccdf(0 | 0,.1);
  target += normal_lpdf(b[6] | 0,.1)
  - 1 * normal_lccdf(0 | 0,.1);
  target += normal_lpdf(b[7] | 0,.1)
  - 1 * normal_lccdf(0 | 0,.1);
  target += normal_lpdf(b[8] | 0,.1)
  - 1 * normal_lccdf(0 | 0,.1);
  target += normal_lpdf(b[9] | 0,.1)
  - 1 * normal_lccdf(0 | 0,.1);
  target += normal_lpdf(b[10] | 0,.1)
  - 1 * normal_lccdf(0 | 0,.1);
  target += normal_lpdf(b[11] | 0,.1)
  - 1 * normal_lccdf(0 | 0,.1);
  target += normal_lpdf(b[12] | 0,.1)
  - 1 * normal_lccdf(0 | 0,.1);
  target += normal_lpdf(b[13] | 0,.1)
  - 1 * normal_lccdf(0 | 0,.1);
  target += normal_lpdf(b[14] | 0,.1)
  - 1 * normal_lccdf(0 | 0,.1);
  target += normal_lpdf(b[15] | 0,.1)
  - 1 * normal_lccdf(0 | 0,.1);
  target += normal_lpdf(b[16] | 0,.1)
  - 1 * normal_lccdf(0 | 0,.1);
  target += normal_lpdf(b[17] | 0,.1)
  - 1 * normal_lccdf(0 | 0,.1);
  target += normal_lpdf(b[18] | 0,.1)
  - 1 * normal_lccdf(0 | 0,.1);
  target += normal_lpdf(b[19] | 0,.1)
  - 1 * normal_lccdf(0 | 0,.1);
  target += normal_lpdf(b[20] | 0,.1)
  - 1 * normal_lccdf(0 | 0,.1);
  target += normal_lpdf(b[21] | 0,.1)
  - 1 * normal_lccdf(0 | 0,.1);
  target += normal_lpdf(b[22] | 0,.1)
  - 1 * normal_lccdf(0 | 0,.1);
  target += normal_lpdf(b[23] | 0,.1)
  - 1 * normal_lccdf(0 | 0,.1);
  target += normal_lpdf(b[24] | 0,.1)
  - 1 * normal_lccdf(0 | 0,.1);
  target += normal_lpdf(b[25] | 0,.1)
  - 1 * normal_lccdf(0 | 0,.1);
  target += normal_lpdf(b[26] | 0,.1)
  - 1 * normal_lccdf(0 | 0,.1);
  target += normal_lpdf(b[27] | 0,.1)
  - 1 * normal_lccdf(0 | 0,.1);
  target += normal_lpdf(b[28] | 0,.1)
  - 1 * normal_lccdf(0 | 0,.1);
  target += normal_lpdf(b[29] | 0,.1)
  - 1 * normal_lccdf(0 | 0,.1);
  target += normal_lpdf(b[30] | 0,.1)
  - 1 * normal_lccdf(0 | 0,.1);
  target += normal_lpdf(b[31] | 0,.1)
  - 1 * normal_lccdf(0 | 0,.1);
  target += normal_lpdf(b[32] | 0,.1)
  - 1 * normal_lccdf(0 | 0,.1);
  target += normal_lpdf(b[33] | 0,.1)
  - 1 * normal_lccdf(0 | 0,.1);
  target += normal_lpdf(b[34] | 0,.1)
  - 1 * normal_lccdf(0 | 0,.1);
  target += normal_lpdf(b[35] | 0,.1)
  - 1 * normal_lccdf(0 | 0,.1);
  target += normal_lpdf(Intercept | 0,.1;
  target += normal_lpdf(sd_1 | 0,.1)
  - 1 * normal_lccdf(0 | 0,.1);
  target += normal_lpdf(z_1[1] | 0, 1);
  target += normal_lpdf(sd_2 | 0,.1)
  - 1 * normal_lccdf(0 | 0,.1);
  target += normal_lpdf(z_2[1] | 0, 1);
  // likelihood including all constants
  if (!prior_only) {
  for (n in 1:N) {
  target += ordered_logistic_lpmf(Y[n] | mu[n], Intercept);
  }
  }
  }
  generated quantities {
  // compute actual thresholds
  vector[ncat - 1] b_Intercept = Intercept + dot_product(means_X, b);
  }

[bgoodri]

@skb

target += normal_lpdf(Intercept | 0,.1;

The right parentheses is missing here in the generated Stan code.

[skb]

Oh! I missed that one. Thank you for your help!