Re: [Gretl-users] Random-effects panel-data
Of course it is a question of how Omega is calculated. The R-program
offers the following options:
plm(y~x,data=datap,model="random",random.method="nerlove")
Model Formula: y ~ x
Coefficients:
(Intercept) x
-1.3212 1.0973
plm(y~x,data=datap,model="random",random.method="walhus")
Model Formula: y ~ x
Coefficients:
(Intercept) x
-1.2388 1.0918
plm(y~x,data=datap,model="random",random.method="amemiya")
Model Formula: y ~ x
Coefficients:
(Intercept) x
-1.2534 1.0928
plm(y~x,data=datap,model="random",random.method="swar")
Error in
swar(object, data, effect) :
the estimated variance of the individual effect is negative
My own programming in R of the formulas in Greene (edition 5) gives
-1.262934 1.09341
My estimate of var(beta) is 0.002613184, whereas the R-alternatives give
0.002514221, 0.002653728, 0.00262924.
Of these, only the first (Nerlove's method of calculating Omega) will
give a positive Hausman test statistic.
Some years ago I had my students do this exercise in GRETL. Then GRETL
gave results that were very similar to mine, except that GRETL gave a
positive Hausman statistic. I suspect that sometimes in the last couple
of years GRETL has gone from estimating Omega with one of
Nerlove/Walus/Amemiya to SWAR (the R default).
Is it possible to choose a method for calculating Omega in GRETL?
best regards
Helgi Tomasson
On Thu, 2012-02-16 at 14:00 -0500, Allin Cottrell wrote:
On Thu, 16 Feb 2012, Riccardo (Jack) Lucchetti wrote:
> On Thu, 16 Feb 2012, Helgi Tomasson wrote:
>
>> I am doing a panel-data exercise from Greene (5th. ed. ) exercise 13.1
>> Why do I get precisely the same estimates in pooled OLS as in the
>> random-effects panel model?
>
> Here's the answer to your question
>
>> 'Within' variance = 3.0455
>> 'Between' variance = 0.113088
>> theta used for quasi-demeaning = 0
>
> In finite samples, it may well happen that the estimate for theta goes
> outside its 'natural' limits (the 0-1 interval), in which case it's
force d
> to 0. See any textbook for details (eg Greene --- used to be section 13.4 in
> the 5th edition, don't known in more recent versions).
Footnote: stata's "xtreg" command produces the same result as
gretl on the data Helgi posted, for the random effects model.
Allin Cottrell
_______________________________________________
Gretl-users mailing list
Gretl-users(a)lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-users