10:21 a.m.
Hello...
While estimating panel model with fixed effects with two explanatory variables for 11
units, number of independent variables in Durbin-Watson test equals 2 or 13? It's
sensible to use this test in case of that kind of models?
Best wishes...
Mariusz
Poland
4:12 p.m.
If you check you will find that Gretl does not include a D-W statistic
in its output for fixed effects panel. The DW statistic is a test
for first order serial correlation. The standard DW statistic is for
a single time series regression. You could possibly design a
corresponding statistic for panel data but it would not be a DW
statistic and it would have a different distribution. If you are
interested in serial correlation and its effects on the variance
covariance matrix you might read the appropriate sections in Cameron
an Trivedi (2005), Microeconometrics, Methods and Applications,
Cambridge.
Best Regards
John
2008/11/27 Mariusz Doszyń <madosz(a)wp.pl>:
Hello...
While estimating panel model with fixed effects with two explanatory
variables for 11 units, number of independent variables in Durbin-Watson
test equals 2 or 13? It's sensible to use this test in case of that kind of
models?
Best wishes...
Mariusz
Poland
_______________________________________________
Gretl-users mailing list
Gretl-users(a)lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-users
--
John C Frain
Trinity College Dublin
Dublin 2
Ireland
www.tcd.ie/Economics/staff/frainj/home.html
mailto:frainj@tcd.ie
mailto:frainj@gmail.com
5:52 p.m.
If you check you will find that Gretl does not include a D-W
statistic
in its output for fixed effects panel.
Yes, it does, at least in polish translation.
The DW statistic is a test
for first order serial correlation. The standard DW statistic is
for
a single time series regression. You could possibly design a
corresponding statistic for panel data but it would not be a DW
statistic and it would have a different distribution. If you are
interested in serial correlation and its effects on the variance
covariance matrix you might read the appropriate sections in Cameron
an Trivedi (2005), Microeconometrics, Methods and Applications,
Cambridge.
Thanks for advices, I'll try to check it!
Best wishes.
Mariusz
Poland
6:58 p.m.
On Thu, 27 Nov 2008, [iso-8859-2] Mariusz Doszyń wrote:
> If you check you will find that Gretl does not include a D-W
> statistic in its output for fixed effects panel.
Yes, it does, at least in polish translation.
True, if there are enough time-series data points. (The DW is
calculated used only the time-series dimension.) But the validity
of this is admittedly questionable.
Allin.
10:33 p.m.
I have found a paper
Bhargava, Franzini, and Narendranathan (1982), Serial Correlation and
the Fixed Effects Model, Review of Economic Studies, Vol 49, No 4,
pp533-549.
For short time dimensions an long unit dimensions the calculate a
Durbin Watson statistic based on the within regression (demeanded
variables). They give upper and lower confidence intervals for T=6
and 10: N = 50, 100, 150, 250, 500 and 1000. and for k = 1, 3, 5, 7,
9, 11, 13 and 17. (T is the time dimension, N is a crosssection
dimension and k is number of variables - the dummies are not included
as they are not in the within equation).
I would still be inclined to think about some form of robust error
estimates or of a dynamic panel analysis. They did not know about
dynamic panel analysis in 1982 and, in any case they did not have the
required computer power.
Best regards
John
2008/11/27 Allin Cottrell <cottrell(a)wfu.edu>:
On Thu, 27 Nov 2008, [iso-8859-2] Mariusz Doszyń wrote:
> > If you check you will find that Gretl does not include a D-W
> > statistic in its output for fixed effects panel.
>
> Yes, it does, at least in polish translation.
True, if there are enough time-series data points. (The DW is
calculated used only the time-series dimension.) But the validity
of this is admittedly questionable.
Allin.
_______________________________________________
Gretl-users mailing list
Gretl-users(a)lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-users
--
John C Frain
Trinity College Dublin
Dublin 2
Ireland
www.tcd.ie/Economics/staff/frainj/home.html
mailto:frainj@tcd.ie
mailto:frainj@gmail.com
5931
days inactive
5931
days old
4 comments
3 participants
participants (3)
-
Allin Cottrell -
John C Frain -
Mariusz Doszyń