This is covered in Godfrey's book (1991, p. 106-107; full ref. below).
Increasing p, where p is the max. power of the fitted values, need not
increase asymptotic local power. For purely spherical errors (when the
test is exact), the choice of p=3 was found to have good power
properties in the simulations of Ramsey & Gilbert (1972). Godfrey &
Orme (1994) suggest the use of p=2 over p=4.
You might prefer the flexibility offered by -ivreset- due to Mark
Schaffer (SSC):
******************
sysuse auto, clear
reg price mpg weight length, robust
ivreset, poly(2)
ivreset, poly(3)
ivreset, poly(4)
******************
--- References ---
@book{godfrey1991misspecification,
title={{Misspecification tests in econometrics: the Lagrange
multiplier principle and other approaches}},
author={Godfrey, LG},
year={1991},
publisher={Cambridge Univ Pr}
}
@article{godfrey1994sensitivity,
title={{The sensitivity of some general checks to omitted variables
in the linear model}},
author={Godfrey, LG and Orme, CD},
journal={International Economic Review},
pages={489--506},
year={1994},
publisher={The Economics Department of the University of
Pennsylvania, and the Osaka University Institute of Social and
Economic Research Association}
}
@article{ramsey1972monte,
title={{A Monte Carlo study of some small sample properties of tests
for specification error}},
author={Ramsey, J. and Gilbert, R.},
journal={Journal of the American Statistical Association},
pages={180--186},
year={1972},
publisher={American Statistical Association}
}
T
On Fri, Sep 11, 2009 at 3:03 PM, Melvyn Weeks <[email protected]> wrote:
> My understanding is that the RESET test in STATA has a default (and
> fixed) setting based upon adding a 4th degree polynomial in predicted
> values.
>
> I can obviously specifiy the test myself.
>
> However, I have noted that using either a robust or non-robust variant,
> that inference can very much depend - especially in small samples - on
> the order of the polynomial.
> Does anyone have experience with this and is there any useful references
> I should check out.
>
> M.
>
> *
> * For searches and help try:
> * http://www.stata.com/help.cgi?search
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/
>
--
To every ω-consistent recursive class κ of formulae there correspond
recursive class signs r, such that neither v Gen r nor Neg(v Gen r)
belongs to Flg(κ) (where v is the free variable of r).
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/