topic Re: Running multiple linear regression with different sets of explanatory variables in SAS Programming

Running multiple linear regression with different sets of explanatory variables

BurakY — Sat, 10 Oct 2020 19:09:21 GMT

Let's assume I have several explanatory variables such as X1, X2, ..., Xn and one dependent variable which is Y.

I want to create regression for all possible paired combinations of X. For example:

y~x1 x2

y~x1 x3

y~x1 x4

...

y~ x10 x11.

What kind of a macro do i need?

Re: Running multiple linear regression with different sets of explanatory variables

PaigeMiller — Sat, 10 Oct 2020 20:25:56 GMT

All possible regressions in SAS

https://support.sas.com/kb/24/986.html

Of course, you asked how to do it, I provide the answer. Is it a good thing to do? Do people actually use this method in general? Well, NO to both questions. If you want to select variables, there are many more effective approaches, including PROC GLMSELECT and PROC PLS.

Re: Running multiple linear regression with different sets of explanatory variables

BurakY — Sat, 10 Oct 2020 20:31:14 GMT

Thanks a lot @PaigeMiller . The macro does even more than I need.

Re: Running multiple linear regression with different sets of explanatory variables

BurakY — Sun, 11 Oct 2020 06:52:08 GMT

Maybe 1 more question. How can I customize this macro to get all possible combination with only 2 independent variables? Because it takes a lot of time macro to run and I don't need combinations with more than 2 independent variables.

Thanks in advance.

Re: Running multiple linear regression with different sets of explanatory variables

PaigeMiller — Sun, 11 Oct 2020 10:50:47 GMT

As I stated, running many regressions is not a good method. I am reluctant to advise you on modifying code from a method that I think is ineffective. Use PROC GLMSELECT to select variables to put into a regression, or use PROC PLS.