how to conduct vectorized summation in SAS/IML

[Sasulee]

Hi all,

I am building a formula which shows below by referring Sum a series in SAS

When I run,

start Sumcon(n);
   i = 1:n;                    /* index of terms */
   return( sum((deltak/k[i]##2)#exp(r[i]#T[i])#Q[i]) );  /* sum of terms */
finish;
n = T(1:68);
sum = j(nrow(n),1);     /* allocate a vector for the results */
do k = 1 to nrow(n);
   sum[k] = Sumcon( n[k] );
end;

It returns ERROR: (execution) Matrix has not been set to a value.

where the variables k, r, t, Q are all (68,1) vectors, the deltak is a fixed value 5.

How can I achieve this formula?

Thanks a lot. 

[PaigeMiller]

ERROR: (execution) Matrix has not been set to a value.

Print out each of the matrices involved and see which one(s) do not have values.

[Sasulee]

Thanks for replying. However, every column is valued. Is it caused by incorrect use of the sum function?

[Rick_SAS]

The vectors are not in scope inside the function. You either need to pass them in or use a GLOBAL statement. See "Understanding local and global variables in the SAS/IML language."

[Rick_SAS · Accepted solution]

Here is how I would compute the sum. I am using 5-dimensional vectors instead of 68 elements:

proc iml;
/* Use n = 5 instead of 68.
   Make up some fake values for the column vectors k, r, t, Q */
k = {1,2,3,4,5};
r = {1,2,3,4,5} / 5;
T = {1,2,3,4,5} / 10;
Q = {1,2,3,4,5} / 15;
deltak = 2;

v = deltak/k##2 # exp(r#T)#Q;
sum = sum(v);

[Sasulee]

Why r, t, Q should divide by 5, 10, 15 respectively? how can I express this when read all var?

[IanWakeling]

Please take another look at Rick's post above, as I think it is the solution that you want. Ignore everything up to deltak=2, as this is artificial data to show that the solution works. Try to include the last 2 lines in your own code, and tell us what happens.

[Rick_SAS]

The comment in my program says "Make up some fake values for the column vectors k, r, t, Q"

Thus, these are fake values. Not real. Made up. Use your values instead.