Thanks to Iwan Barankay who suggested that I could run a two-way error
component fixed effect model by use of:-
xi: reg y i.individual i.year x
I've checked this with my data and with my balanced panel it looks
right, with just one small problem, that the sum of the dummies on each
of the groups no longer sums to zero. My manual solution is to divide
the sum of the dummies' coefficient values add this to the overall
intercept term whilst subtracting this sum divided by the number of
individuals within the group to each of the individual dummy
coefficients.
I think this is right, but is there anyway to do this automatically?
Chris Bojke wrote:
>
> Has anyone any idea whether there exists a Stata ado command which can
> estimate a two-way error component regression model for panel data
> (where the first error is often characterised as an individual effect
> and the second as a time effect which applies to all individuals at that
> discrete time)? I'm particularly interested in applying this for a
> fixed-effects model.
>
> I've net searched and read the statalist archives to no avail. Quite
> surprised given that the model appears frequently in textbooks.
begin:vcard
n:Bojke;Chris
tel;cell:0795 818 3140
tel;fax:(01904) 432700
tel;work:(01904) 432694
x-mozilla-html:FALSE
org:National Primary Care Research & Develpoment Centre;Centre for Health Economics
version:2.1
email;internet:[email protected]
title:Research Fellow
adr;quoted-printable:;;Room D146=0D=0AAlcuin College=0D=0AUniversity of York;York;;YO10 5DD;United Kingdom
fn:Chris Bojke
end:vcard
