topic Re: proc IML in SAS/IML Software and Matrix Computations

Regression in proc IML

Golf — Mon, 30 Oct 2023 10:34:07 GMT

Hello, I have a model;
Y = b0+ b1X1+b2X2 + u
I want to use proc IML to create a column vector:
[beta] = [1 -b0 b1 -b2]'. [Dimension: 4x1]
I also want to create a matrix of X
[X ] = [Y ones X1 X2]. [Dimension: nx4]
Then I calculate the vector of coint:
coint = [X]*[beta]. [Dimension: nx1]

I started with the following codes;

proc reg data = a outest = coefficients;
model y = x1 x2;
run;

How can I use Proc IML to obtain the desired results?

Thank You

Re: proc IML

sbxkoenk — Mon, 30 Oct 2023 09:57:28 GMT

I have moved this topic to 'SAS/IML' - board.

SAS/IML 14.2 User’s Guide
Tutorial: A Module for Linear Regression
A Module for Linear Regression
https://go.documentation.sas.com/doc/en/imlcdc/14.2/imlug/imlug_tutorial_sect003.htm

Home > Analytics > SAS/IML > Multiple linear regression using proc iml
https://communities.sas.com/t5/SAS-IML-Software-and-Matrix/Multiple-linear-regression-using-proc-iml/td-p/844255

BR,
Koen

Re: proc IML

Rick_SAS — Mon, 30 Oct 2023 10:51:27 GMT

I don't understand why there are some (but not all) minus signs in the regression coefficient, but I think this program should get you started. If you intended to use -b1, then insert it in the obvious location.

proc iml;
/* use sample data */
use sashelp.class;
read all var "Weight" into Y;
read all var {"Height" "Age"} into XVars;

/* specify regression coefficients */
b0 = 13;
b1 = 9;
b2 = 5;

/* [beta] = [1 -b0 b1 -b2]'.      [Dimension: 4x1] */
beta = 1 // -b0 // b1 // -b2;

/* [X ] = [Y ones X1 X2].          [Dimension: nx4] */
ones = j(nrow(Y), 1, 1);
X = Y || ones || XVars;
coint = X*beta;
print Y coint;

Re: proc IML

Golf — Mon, 30 Oct 2023 12:45:00 GMT

Thank You. @Rick_SAS

Re: proc IML

Golf — Mon, 30 Oct 2023 12:45:20 GMT

Thank You.

Re: proc IML

Rick_SAS — Fri, 12 Jan 2024 14:02:29 GMT

Is there anything else we can help with? If your question is answered, please mark the solution and close this thread.