Checking for equality between a vector element and a real

General
1

I have a vector like

vector[100] A

and I want to check if the ith element is equal to 0

if(A[i] == 0.0) { do stuff }

This generates a compilation error

SYNTAX ERROR, MESSAGE(S) FROM PARSER:
Binary infix operator == with functional interpretation logical_eq requires arguments of primitive type (int or real), found left type=vector, right arg type=real.
No matches for:

logical_eq(vector, real)

Available argument signatures for logical_eq:

logical_eq(int, int)
logical_eq(int, real)
logical_eq(real, int)
logical_eq(real, real)

Conditions in if-else statement must be primitive int or real; found type=ill-formed
error in ‘model11c114eb458a_45c192425899c18b7d0d3f183a06b147’ at line 90, column 9

How do I go about doing this in Stan?

2

This model compiles for me:

parameters {
  vector[100] A;
}

model {
  int i = 1;
  if(A[i] == 0.0) {
    A ~ normal(0, 5);
  } else {
    A ~ normal(0, 1);
  }
}

Can you check that this compiles for you? Is there anything different with the model you’re testing?

3

Yes, it compiles. Here is a trimmed-down version of the code I am trying to compile. The goal is to subset the vector M as part of an effort to generate a prior predictive distribution. I am running rstan Version 2.21.2 on R V4.0.3, all on Linux Mint 20.

test <- "

functions {

int count_urn(int[] sub,vector[] dT,int index) {
  int count;
  count = 0;
  for(i in 1:num_elements(sub)) {
    if(sub[i] == index) {
      if(dT[i] > 0.0) {
        count = count + 1;
      }
    }
  }
  return(count);
}

int[] indu(int[] sub,vector[] dT,int index) {
  int res[count_urn(sub,dT,index)];
  int ci;
  ci = 1;
  for(i in 1:num_elements(sub)) {
    if(sub[i]==index) {
      if(dT[i] > 0.0) {
        res[ci] = i;
        ci = ci + 1;
      }
    }
  }
  return(res);
}
} 
  
data {
  int sub[100];
  vector[100] dT;
} 

generated quantities{
  vector[100] M;
  M[indu(sub,dT,11)];
}
"

stan_model(model_code=test,verbose=TRUE)

The error issued by the compiler is

TRANSLATING MODEL '93da0f9f7e830d4636774afc0967dc4e' FROM Stan CODE TO C++ CODE NOW.
SYNTAX ERROR, MESSAGE(S) FROM PARSER:
Binary infix operator >= with functional interpretation logical_gt requires arguments of primitive type (int or real), found left type=vector, right arg type=real.
No matches for: 

  logical_gt(vector, real)

Available argument signatures for logical_gt:

  logical_gt(int, int)
  logical_gt(int, real)
  logical_gt(real, int)
  logical_gt(real, real)

Conditions in if-else statement must be primitive int or real; found type=ill-formed
 error in 'modeledf5fe5354e_93da0f9f7e830d4636774afc0967dc4e' at line 10, column 9
  -------------------------------------------------
     8:   for(i in 1:num_elements(sub)) {
     9:     if(sub[i] == index) {
    10:       if(dT[i] > 0.0) {
                ^
    11:         count = count + 1;
  -------------------------------------------------

PARSER EXPECTED: <expression>
Error in stanc(file = file, model_code = model_code, model_name = model_name,  : 
  failed to parse Stan model '93da0f9f7e830d4636774afc0967dc4e' due to the above error.
4

The issue here is that you’ve declared the input dT as an array of vectors (vector[]). This means that dT[i] is selecting the i-th vector from the array, rather than the i-th element of a vector. If dT is supposed to be a single vector, then you just need to change the declaration to: int count_urn(int[] sub,vector dT,int index)

(i.e., vector instead of vector[])

3 Likes