Page 1 of 1
areg
Posted: Tue Apr 05, 2016 7:09 pm
by CharlieEVIEWS
Can I suggest something similar to Stata's areg in EViews (or to increase the default size of C)? I keep bumping into the wall using @expand on a categorical with a large number of categories using the default coef vector and i really dont want to estimate an equation by expression with 1500 categories with a custom coeff vector.... the only real feasible solution I can think of is to COM out to matlab, which isn't ideal (maybe there's a better way however?)
Re: areg
Posted: Tue Apr 05, 2016 7:23 pm
by EViews Gareth
What does areg do?
You don't need to estimate by expression to use a custom coef vector. You can estimate by list then change the name of the coefficient vector on the options tab.
Re: areg
Posted: Tue Apr 05, 2016 7:44 pm
by CharlieEVIEWS
Even when using @expand? I keep getting "Too many index series in @expand" even when using a custom set coef. In stata, areg 'absorbs' categorical variables with a large number of categories, producing estimates for explanatory variables but not category (dummy) coefficients.
Re: areg
Posted: Tue Apr 05, 2016 9:54 pm
by EViews Gareth
Code: Select all
wfcreate u 10000
series i = @floor(rnd*2000)
series y=nrnd
coef(2100) beta
equation eq1.ls(coef=beta) y @expand(i)
That has 2000 coefficients and works just fine.
Re: areg
Posted: Tue Apr 05, 2016 9:54 pm
by startz
Yeah. Sometimes you want fixed effects taken out but really don't have a panel structure.
Re: areg
Posted: Wed Apr 06, 2016 6:08 am
by CharlieEVIEWS
Sorry -- was expanding a group which was already expanded.
.
However, still cheering for an AREG option!
Re: areg
Posted: Wed Apr 06, 2016 12:31 pm
by EViews Glenn
Noted.
You can, of course, to define an undated panel using the categorical variable as the ID series. That will provide you with access to all of the tools of panels, including clustered standard errors.