SAS PROC IML AND SQL under precision matrix product

[LineMoon]

Hello,

I would like to konw which  method give the best result for the matrix with the number between 0 et 1:

==>  I have two matrix A and B,  0<=Aij<=1,  0<= Bij<=1

Which metho gives the best results :

1- IML Proc

or

2- proc sql

do i=1 to L;

proc sql;
create table prod as
select (...)+(....) as  Vi
t1 m ,Tran_t2 n;
quit;

end;

 Any documentation for that are welcomed

[Rick_SAS]

For matrix computations, use IML.

Your title says "matrix product," by which I assume you want the standard matrix multiplication:

C = A*B;

However, read this article "Ways to multiply in the SAS/IML language" if you want a different product.

[LineMoon]

Thank you for your answer.

Yes, I want to do C=A*B

But, I can use IML proc or SQL proc , I have a small number.

Please, which method (IML or SQL) is better for the small number ?

[Rick_SAS]

As I said, you should  use IML for matrix computations.

All SAS procedures use double precision computations. If your numbers are so small that you think the computation will result in a numerical overflow or underflow, it probably doesn't make a difference how you perform the computation. If you post the data and a sample program, we might be able to say more.

[LineMoon]

Thank you for your message.

As an exemple,

You can take

A= (Aij)==> 0<Aij<1

B = (Bij)==> 0<Bij<1

Methode 1: Using IML proc

C=A*B

Method  2: Using SQl poc

C= A*transpose(B)

[Ksharp]

For the precise problem , you'd better use IML's function XMULT() to calculate Matrix Product.

z = xmult(x,y); 
print z;

Look like FORMAT is not good for displaying precision. 

[LineMoon]

Thank you for your answer.

It is a good idea.

What's about  the results found by sql proc ? C=A*transpose(B)