Statalist The Stata Listserver

[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: Re: Newey estimations

From   "Rodrigo A. Alfaro" <>
To   <>
Subject   st: Re: Newey estimations
Date   Wed, 2 Aug 2006 22:08:53 -0400


You don't have likelihood with Newey-West procedure. The
procedure is OLS but with 'corrected'-standard errors (SE). 
The SE are computed in a second step. Then you cannot 
compute LR test. Also for the case of one coefficient (see 
Clive's example) F is the same as t^2, then a huge value in F
implies a small SE.

I haven't read Andrews' paper, but I think that the results there
are not applicable to newey's regressions. My point is based on
the fact that you are not choosing the optimal lag jointly with the
structural break (even if you know that the number of breaks!).
This point will change the asymptotic distributions, which I guess
these are the results in Andrews' paper. 

I am not strong in "structural-breaks" but I know that there is
not an easy task. I suggest you to read a survey prepared by
Pierre Perron "Dealing with Structural Breaks" available in his
webpage, for sure he is an authority
in the subject and you will find the right tool for your problem.

Good luck,

----- Original Message ----- 
From: "Evelyn Colino de Cantero" <>
To: <>
Sent: Wednesday, August 02, 2006 7:25 PM
Subject: st: Newey estimations


I'm running a newey-west regression with command "newey".
I want also to test for structural break. I'm using the
optimal tests suggested by Andrews and Ploberger (1994) - the
Mean-Wald test - 

as part of the test I need to compute the standard LR test -
or a F-test for a parameter's significance (the one measuring
a possible change in the coeff.)

If I run the command "test" after "Newey" it shows a F-test
result, (which is actually really huge compared with the same
test under a regular OLS regression on the same equation)

Does somebody knows what is the exactly procedure Stata
follows to test parameters under linear reg. with Newey-West
std. errors estimators?
Is still valid to use the "test" command for this pourpose?

I'll appreciate any suggestions,

Evelyn Colino
PhD Student
University of Illinois at U-C
Research Assistant Agricultural and Consumer Economics Department
326 Mumford Hall, MC-710
1301 West Gregory Drive
Urbana, Il., 61801-3605

*   For searches and help try:

*   For searches and help try:

© Copyright 1996–2021 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index