Real return type required error for binomial

General
1

Hi everybody.
I a new in using Stan. I searched in the FAQ but did not find a solution for my problem.
My model is not very complicated, but I have an error I do not understand. In my model block, when I specify the likelihood, it seems that the binomial function requires a real. The delta parameter is for me a vector of real numbers, so I do not understand the problem.

data {
  int<lower=1> N;
  vector[N] n0;
  vector[N] n;
}

// The parameters of the model. 
parameters {
  vector<lower=0>[N] lambda;
}
// Transformed parameters. Here theta and delta calculated from lambda
transformed parameters {
  vector[N] theta;    
  vector[N] delta; 
  theta = cumulative_sum(lambda);
  delta = exp(-theta);
}
// The model to be estimated. 
model {
  vector[N] logitlambda;
  real temp;
  logitlambda = logit(lambda);
  // prior
  for (i in 1:61){
      logitlambda[i] ~ normal(0, 1000);
  } 
  temp ~ normal(0, 1000);
  for (i in 62:N){
      logitlambda[i] = temp;
  }  
// likelihood
  for (i in 1:N){
      n0[i] ~ binomial(n[i], delta[i]);  // here the error, with the index i of delta underlined with a tilde
  }
}

Should I use an array instead of a vector for delta? But I think it would be less efficient.
I also tried to replace the loop of the likelihood by

n0 ~ binomial(n,delta);
``
But I still have the same error with delta underlined.

Any help would be appreciated.
2

The issue is not with delta but with n0 and n. They need to be integer types.

You can change your data block to contain

  array[N] int n0;
  array[N] int n;

or the old syntax for arrays (only recommended if you are using an older Stan version such as RStan on CRAN)

  int n0[N];
  int n[N];

After doing this you can also replace the for loop with the vectorized version you mention in the post

3

Thanks a lot! Changing the data block as suggested and replacing the loop with the vectorized version seems to be OK, as nothing is underlined in red tildes now in my model.
However, I have an error when trying to fit the model:
PARSER FAILED TO PARSE INPUT COMPLETELY
STOPPED AT LINE 1:
I am sure it should be something silly I am missing because I am new in Stan, but I really don’t see what can be the problem with line 1…

I exactly paste my model here, with all the lines and comments:

//
// The input data: 
// 'n0' an array of integers,
// 'n' an array if integers, 
// 'n0' and 'n'  of length 'N'.
data {
  int<lower=1> N;
  int<lower=0> n0[N];
  int<lower=0> n[N];
}

// The parameters of the model.
parameters {
  vector<lower=0>[N] lambda;
}
// Transformed parameters. Here theta and delta calculated from lambda
transformed parameters {
  vector[N] theta;    
  vector[N] delta; 
//  real delta[N];
  theta = cumulative_sum(lambda);
  delta = exp(-theta);
}
// The model to be estimated. 
model {
  vector[N] logitlambda;
  real temp;
  logitlambda = logit(lambda);
  // prior
  for (i in 1:61){
      logitlambda[i] ~ normal(0, 1000);
  } 
  temp ~ normal(0, 1000);
  for (i in 62:N){
      logitlambda[i] = temp;
  }  
  n0 ~ binomial(n,delta);
}

I specify that I have an error when using the fit function:

fit = stan(
  model_code = "Model2.stan",
  data   = dat,
  iter   = 200,
  warmup = 100,
  chains = 4,
  seed=4938483
)

If I used stan_model(file=“Model2.stan”), I don’t have an error.

4

I believe your model code is fine, but its possible your data is invalid (either the wrong shape or possibly you have real numbers in n or n0)

5

Thanks for your answer. I tried different ways to specify my input data, but I always have the same error when fitting the model: PARSER FAILED TO PARSE INPUT COMPLETELY
STOPPED AT LINE 1: Model2.stan
Here is my input data, specifying that n0 and n are arrays of integers as expected by the Model2.stan

n0 <- array(as.integer(c(6,4,5,6,6,1,6,4,8,1,6,9,5,8,10,7,10,7,22,11,22,52,34,22,23,22,30,31,30,32,34,34,39,17,27,24,26,40,16,22,22,30,25,21,20,22,24,22,22,21,26,27,10,22,14,17,16,11,23,5,25,33,9,16,7,13,8,6,3,9,11,13,4,5,4,8,1,5,4,0,3,6,1,0,2,1,0,0,1,0,0,1,0,0,1)))
n <- array(as.integer(c(6,4,5,6,6,1,6,4,8,1,6,9,5,8,1,7,10,7,22,11,22,52,34,22,23,22,30,31,30,32,34,34,39,17,27,24,26,40,16,22,22,30,25,21,20,22,24,22,22,21,26,27,10,22,14,17,16,11,23,5,25,33,9,16,7,13,8,6,3,9,11,13,4,5,4,8,1,5,4,0,3,6,1,0,2,1,0,0,1,0,0,1,0,0,1)))

dat = list(
  N = length(n0),
  n0 = n0,
  n = n
)

And the error is when running:

fit = stan(
  model_code = "Model2.stan",
  data   = dat,
  iter   = 200,
  warmup = 100,
  chains = 4,
  seed=4938483
)