Help in finding solutions for the error: 0,248,Js_of_ocaml__Js.Error,18,TypeError: not a function Error in model_cppcode$errors : $ operator is invalid for atomic vectors

Modeling
,
1

Here is the program.

set.seed(123)
AF <- readRDS("AF_clean.rds")
N= nrow(AF)
AF[] <- lapply(AF, function(x) if(is.numeric(x)) as.integer(x) else x)

#################################################################################
x2<-AF$x2
x3<-AF$x3
x4<-AF$x4
x5<-AF$x5
eta1<-AF$eta1
eta2<-AF$eta2
eta3<-AF$eta3
eta4<-AF$eta4
eta5<-AF$eta5
eta6<-AF$eta6
eta7<-AF$eta7
sim_ordinal <- function(n, probs) {
  cumprobs <- cumsum(probs)
  u <- runif(n)
  sapply(u, function(x) sum(x > cumprobs) + 1)
}

eta8 <- as.integer(sim_ordinal(N, c(0.35, 0.35, 0.3)))
group <- rbinom(N, 1, 0.5)
x6 <- as.numeric(ifelse(group == 0,
             rlnorm(N, meanlog = 4.24, sdlog = 0.32),
             rlnorm(N, meanlog = 4.77, sdlog = 0.18)))
logit <- -2 + 0.5*AF$eta1 + 0.3*eta8 + 0.4*AF$eta7
xi <- rbinom(N, 1, plogis(logit))

stan_data_AF <- list(
  N = nrow(AF), 
  x2 = x2,
  x3 = x3,    
  x4 = x4,
  x5 = x5,   
  x6 = x6,    
  xi = xi,    
  
  eta1 = eta1,
  eta2 = eta2,
  eta3 = eta3,
  eta4 = eta4,
  eta5 = eta5,
  eta6 = eta6,
  eta7 = eta7,
  eta8 = eta8
)


stan_code_AF <- "
data {
  int<lower=0> N;  // number of observations
  int<lower=1,upper=5> eta1[N];
  int<lower=1,upper=5> eta2[N];
  int<lower=1,upper=3> eta3[N];
  int<lower=1,upper=5> eta4[N];
  int<lower=1,upper=5> eta5[N];
  int<lower=1,upper=5> eta6[N];
  int<lower=1,upper=5> eta7[N];
  int<lower=1,upper=3> eta8[N];
  int<lower=1,upper=1> x2[N];
  int<lower=0> x3[N];
  int<lower=0> x4[N];
  int<lower=0,upper=1> x5[N];
  real<lower=0> x6[N];
  int<lower=0,upper=1> xi[N];
}

parameters {
  real beta_eta1_eta2;
  real beta_eta2_eta3;
  real beta_eta3_eta4;
  real beta_eta1_xi;
  real beta_eta1_eta4;
  real beta_x2_eta2;
  real beta_x3_eta4;
  real beta_eta5_eta3;
  real beta_eta6_eta3;
  real beta_eta3_eta8;
  real beta_x6_eta8;
  real beta_x4_eta8;
  real beta_x5_eta8;
  real beta_eta8_xi;
  real beta_eta7_xi;

  real alpha_eta2;
  real alpha_eta3;
  real alpha_eta4;
  real alpha_eta8;
  real alpha_xi;

  real<lower=0> sigma_eta2;
  real<lower=0> sigma_eta3;
  real<lower=0> sigma_eta4;
  real<lower=0> sigma_eta8;
}

model {

  beta_eta1_eta2 ~ normal(0, 1);
  beta_eta2_eta3 ~ normal(0, 1);
  beta_eta3_eta4 ~ normal(0, 1);
  beta_eta1_xi ~ normal(0, 1);
  beta_eta1_eta4 ~ normal(0, 1);
  beta_x2_eta2 ~ normal(0, 1);
  beta_x3_eta4 ~ normal(0, 1);
  beta_eta5_eta3 ~ normal(0, 1);
  beta_eta6_eta3 ~ normal(0, 1);
  beta_eta3_eta8 ~ normal(0, 1);
  beta_x6_eta8 ~ normal(0, 1);
  beta_x4_eta8 ~ normal(0, 1);
  beta_x5_eta8 ~ normal(0, 1);
  beta_eta8_xi ~ normal(0, 1);
  beta_eta7_xi ~ normal(0, 1);
 
  alpha_eta2 ~ normal(0, 2);
  alpha_eta3 ~ normal(0, 2);
  alpha_eta4 ~ normal(0, 2);
  alpha_eta8 ~ normal(0, 1);
  alpha_xi ~ normal(logit(0.29), 2.5);
  
  sigma_eta2 ~ cauchy(0, 2.5);
  sigma_eta3 ~ cauchy(0, 2.5);
  sigma_eta4 ~ cauchy(0, 2.5);
  sigma_eta8 ~ cauchy(0, 2.5);

  // Likelihood
  for (n in 1:N) {
    // Structural equations
      real mu_eta2;  
      real mu_eta3;
      real mu_eta4;
      real mu_eta8;
      real logit_xi;
      
      mu_eta2 = alpha_eta2 + beta_eta1_eta2 * eta1[n] + beta_x2_eta2 * x2[n];
      mu_eta3 = alpha_eta3 + beta_eta2_eta3 * eta2[n] + beta_eta5_eta3 * eta5[n] + 
                   beta_eta6_eta3 * eta6[n];
      mu_eta4 = alpha_eta4 + beta_eta3_eta4 * eta3[n] + beta_eta1_eta4 * eta1[n] + 
                   beta_x3_eta4 * x3[n];
      mu_eta8 = alpha_eta8 + beta_eta3_eta8 * eta3[n] + beta_x6_eta8 * x6[n] + 
                   beta_x4_eta8 * x4[n] + beta_x5_eta8 * x5[n];

    // Binary outcome
      logit_xi = alpha_xi + beta_eta1_xi * eta1[n] + beta_eta8_xi * eta8[n] + 
                    beta_eta7_xi * eta7[n];

    // Likelihood contributions
      eta2[n] ~ normal(mu_eta2, sigma_eta2);
      eta3[n] ~ normal(mu_eta3, sigma_eta3);
      eta4[n] ~ normal(mu_eta4, sigma_eta4);
      eta8[n] ~ normal(mu_eta8, sigma_eta8);
      xi[n] ~ bernoulli_logit(logit_xi);
  }
}

generated quantities {
  vector[N] log_lik;
  for (n in 1:N) {
    real mu_eta2 = alpha_eta2 + beta_eta1_eta2 * eta1[n] + beta_x2_eta2 * x2[n];
    real mu_eta3 = alpha_eta3 + beta_eta2_eta3 * eta2[n] + beta_eta5_eta3 * eta5[n] + 
                   beta_eta6_eta3 * eta6[n];
    real mu_eta4 = alpha_eta4 + beta_eta3_eta4 * eta3[n] + beta_eta1_eta4 * eta1[n] + 
                   beta_x3_eta4 * x3[n];
    real mu_eta8 = alpha_eta8 + beta_eta3_eta8 * eta3[n] + beta_x6_eta8 * x6[n] + 
                   beta_x4_eta8 * x4[n] + beta_x5_eta8 * x5[n];
    real logit_xi = alpha_xi + beta_eta1_xi * eta1[n] + beta_eta8_xi * eta8[n] + 
                    beta_eta7_xi * eta7[n];

    log_lik[n] = normal_lpdf(eta2[n] | mu_eta2, sigma_eta2) +
                 normal_lpdf(eta3[n] | mu_eta3, sigma_eta3) +
                 normal_lpdf(eta4[n] | mu_eta4, sigma_eta4) +
                 normal_lpdf(eta8[n] | mu_eta8, sigma_eta8) +
                 bernoulli_logit_lpmf(xi[n] | logit_xi);
  }
}
"
fit_AF <- stan(model_code = stan_code_AF, data = stan_data_AF,
+                  chains = 4, 
+                  iter = 5000,    # Increased from 2000
+                  warmup = 1500,  # Increased from 500
+                  control = list(adapt_delta = 0.99)) 

0,248,Js_of_ocaml__Js.Error,18,TypeError: not a function
Error in model_cppcode$errors : $ operator is invalid for atomic vectors

> class(model_cppcode)
Error: object 'model_cppcode' not found

[edit: escaped code and run]

2

Hi, @liatchin and welcome to the Stan forums.

The error you’re getting is coming from our OCaml parser and code generator—that’s the component that converts a Stan program to C++ code. If I try compiling your Stan program with the latest version, I run into a different issue.

--- Translating Stan model to C++ code ---
bin/stanc --filename-in-msg=liatchin.stan --o=/Users/carp/temp2/liatchin.hpp /Users/carp/temp2/liatchin.stan
Error in 'liatchin.stan', line 16, column 2: Declaration of arrays by placing
    brackets after a variable name was removed in Stan 2.33.0. Instead use
    the array keyword before the type. This can be changed automatically
    using the auto-format flag to stanc

That’s because RStan is still using our old array syntax—we’ve ben blocked on updating it on CRAN. CmdStanR is up to date, but would require rewriting the syntax.

Can you run simpler models in the install you have?

I’m not sure where that issue’s coming from in OCaml. Which platform are you running on (Windows, Mac OS, Linux)? And which version of R? I’m pinging @WardBrian, who knows the OCaml code, in case he knows where this comes from.

3

These issues have come up a few times recently, and we believe they are actually bugs in the underlying JS engine that the compiler code is just running afoul of. @andrjohns knows more, I think.

I believe we have had success by either having users install the V8 R package and/or installing the newer version of RStan available from r-universe