topic Re: How to enter multiple variables in the linear regression model to select predictors ?? NOT A SPA in New SAS User

How to enter multiple variables in the linear regression model to select predictors ?? NOT A SPAM !!

brook2 — Sat, 13 Apr 2019 21:05:25 GMT

I have a data set with 140 rows and 81 variables including the dependent variable, I want to include all of the variables to do backward elimination for designing regression model. How do i enter all of the variables at once. I found a similar question here in the following link but the answer is to vague to reproduce in my own data set

proc glm data = train plot ;
model SalePrice = GrLivArea ........ 80 variables ;
run;

what if there is categorical variable , how do i address that when selecting predicting worthiness , binary coding may help but what if 0 is selected and not 1 in the selection criterion. The data also have almost 24 variables char data type and the rest are numeric. How do I address the problem as

proc glmselect data=TrainSubset;

will give an error saying

ERROR: Variable BsmtFinType1 should be either numeric or specified in the CLASS statement

I have found same discussion thread here but found it to vague and complicated

https://blogs.sas.com/content/iml/2017/02/13/run-1000-regressions.html

Thank you !!!

Re: How to enter multiple variables in the linear regression model to select predictors ?? NOT A SPA

PaigeMiller — Sat, 13 Apr 2019 21:40:06 GMT

proc glmselect data=TrainSubset;

will give an error saying

ERROR: Variable BsmtFinType1 should be either numeric or specified in the CLASS statement

You haven't specified BsmtFinType1 properly, as the error message states. If it is numeric (is it?) then you need to take it out of the CLASS statement. If it is categorical (is it?), then you need to put it in the CLASS statement.

I have a data set with 140 rows and 81 variables including the dependent variable, I want to include all of the variables to do backward elimination for designing regression model.

Go to Google and type in "problems with stepwise regression", and read some of those articles. Then, don't do stepwise or backwards or forwards variable selection.

Consider doing Partial Least Squares Regression on this data (PROC PLS in SAS), as PLS is known to have better properties than any form of stepwise on data like this.

Re: How to enter multiple variables in the linear regression model to select predictors ?? NOT A SPA

Ksharp — Sun, 14 Apr 2019 11:11:35 GMT

Best way is typing these variables one by one .

Or try this one .I am not sure if it work.

'SalePrice ' must be in the first position.

proc glm data = train plot ;

class _character_ ;
model SalePrice = First_X_Variable -- Last_X-Variable ;
run;