EViews.com

Posted: **Sat Jun 25, 2016 12:47 pm**

Dear all

I want to run a stepwise least squares regression (unidirectional backwards). This seems straightforward in EVIEWS. Just select equation, method: STEPS, fill in your variables and adjust the options.

Problem: I have 250+ variables. I want to regress one variable on its first two lags and on the remaining variables and 1st order lags of those variables. My question is:is there a quick way to do this? Doing this manually would be pretty much insane and I assume there has to be another way.

Other question: does EVIEWS have commands for lasso and boosted regression trees for model selection?

Thanks!

Posted: **Sat Jun 25, 2016 1:34 pm**

Add all 250 series to a group. Then write a quick program that adds those series plus their lags to another group. Then use that group in your estimation.

No LASSO or boosted reg.

Code: Select all


group group2
for !i=1 to group1.@count
%name = group1.@seriesname(!i)
group2.add {%name} {%name}(-1)
next

EViews.com

model selection: steps, lasso, boosted regression trees

model selection: steps, lasso, boosted regression trees

Re: model selection: steps, lasso, boosted regression trees