Page 1 of 1

Stepwise backward with nvars=x

Posted: Tue Nov 13, 2012 5:13 am
by Patrick

Code: Select all

stepls(method=uni,nvars=4)
Works as expected.

Code: Select all

stepls(method=stepwise,nvars=4)
Works as expected as well.

Code: Select all

stepls(method=uni,nvars=4,back)
or

Code: Select all

stepls(method=stepwise,nvars=4,back)
Gives me more than 4 variables but seems to stop on the criterion on p<0.5 regardless of what i set nvars to. Can I run backwards with exactly 4 variables no matter what p or t I get?

Thanks is advance guys.
/Patrick

As a side note Eviews spelled search wrong in "Stopping criterion: number of seach regressors in..." when you go uni-directional.

Re: Stepwise backward with nvars=x

Posted: Tue Nov 13, 2012 8:45 am
by EViews Gareth
You cannot force it to have 4 regressors when going backwards.

Thanks for the typo note.

Re: Stepwise backward with nvars=x

Posted: Tue Nov 13, 2012 9:15 am
by Patrick
Thanks Gareth. May I ask why we can not force nvars when going backwards?

Re: Stepwise backward with nvars=x

Posted: Tue Nov 13, 2012 9:34 am
by EViews Gareth
Because the other stopping criteria will kick in first. Why not use combinatorial?

If you really want to use backwards stepwise, you can artificially set the p-value criteria so that they are unimportant (which is the same as doing uni-directional backwards):

Code: Select all

STEPLS(FTOL=0.000001,BTOL=0,NVARS=4,BACK)