Page 1 of 1

Newey West standard errors in Stata vs Eviews

Posted: Mon Feb 14, 2011 2:29 pm
by oxidas
Hello,

I have recently switched to Eviews 7 and one of the things I noticed is that they the procedures for estimating timeseries with NW standard errors are different. In Stata you would simply type "Newey and specify lag length", whereas as I understand that in Eviews you cannot manually choose the lag length and you also have to specify whitening options, kernel choice,etc.

I was wondering what would be the settings for whitening, kernel and other options to get the same or as close as possible results to the method used by Stata?

Re: Newey West standard errors in Stata vs Eviews

Posted: Mon Feb 14, 2011 2:50 pm
by EViews Gareth
Stata is somewhat fragmented. Which particular Stata command were you using?

Re: Newey West standard errors in Stata vs Eviews

Posted: Tue Feb 15, 2011 12:36 am
by oxidas
I was using Newey (depvar) (indepvars), lag(x)

Re: Newey West standard errors in Stata vs Eviews

Posted: Tue Feb 15, 2011 8:54 am
by EViews Gareth
ok. The equivalent EViews command would be:

Code: Select all

LS(COV=HAC,COVBW=Z) (depvar) (indepvars)
where Z is X+1 (i.e. the user lag-length in EViews is the same as the Stata one plus one).