Loops for multi-dimensional arrays

Modeling
1

Hi everybody!

I have an issue with the part of the code below. I define three-dimensional array “pricing_kernel” and vector “beta”. I would like to run the loop and write the expression “exp(pricing_kernel[N,2,j]*beta)” to the corresponding shelf of the object “mm”.

data {
int<lower=0> N;             
int<lower=0> nq;             
int<lower=0> J;                        

real pricing_kernel[N,2,J]; 
}

parameters {
vector[2] beta;
}

model {

real mm[N,1,J];


//priors:
beta[1]~normal(0,1);
beta[2]~normal(-2,0.5);

for (j in 1:J) {    
   
   mm[N,1,j] = exp(pricing_kernel[N,2,j]*beta) ;  
 }
.....
}

In R it can be made easy enough:

  for (j in 1:J) {        
    mm[,,j] = exp(pricing_kernel[,,j] %*% beta );  
  }

I tried to do it in Stan in different ways, but not successfully. Could someone help with this? Thank you in advance!

2

Hi there!

The dimensions of your multiplication don’t quite make sense. I’d guess that you receive an error on

mm[N,1,j] = exp(pricing_kernel[N,2,j]*beta) ;

because here you are multiplying a real scalar (which is pricing_kernel[N,2,j]) with a 2-dimensional vector (beta). The result of this is a 2-dimensional vector. Then you take the exp function, which (I believe) again results in a 2-dimensional vector (the element-wise exp of the previous vector). Finally, you try to assign this 2-dimensional vector to a real scalar, mm[N,1,j].

What exactly are you trying to accomplish? If you want to accomplish the same multiplication as you have in R:

for (j in 1:J) {        
  mm[,,j] = exp(pricing_kernel[,,j] %*% beta );  
}

then the easiest way is probably to declare pricing_kernel as an array of matrices:

matrix[N,2] pricing_kernel[J];

then pricing_kernel[j] would be a matrix for every j in 1:J.

1 Like
3

Thank you very much for your response!
Yes, I would like to accomplish the same multiplication as in R. I had tried declaring “pricing_kernel” as you suggested, but it didn’t work out.

matrix[N,2] pricing_kernel[J];  
matrix[N,1] mm[J];
 for (j in 1:J) {    
 
   mm[j] = exp(pricing_kernel[j]*beta) ;  
 }

When I run the script, it returns the ERROR message:

"SYNTAX ERROR, MESSAGE(S) FROM PARSER:
Dimension mismatch in assignment; variable name = mm, type = matrix; right-hand side type = vector.
Illegal statement beginning with non-void expression parsed as
mm[j]
Not a legal assignment, sampling, or function statement. Note that

  • Assignment statements only allow variables (with optional indexes) on the left;
  • Sampling statements allow arbitrary value-denoting expressions on the left.
  • Functions used as statements must be declared to have void returns…"
4

Please try declaring mm as

vector[N] mm[J];

and let me know if that works.

Otherwise you might need to play around with the to_matrix function (see HERE ).

5

Thanks! It seems it works now!

1 Like
6

Great! Glad I could help :)

1 Like