Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: Re: xtreg v areg


From   "Clive Nicholas" <Clive.Nicholas@newcastle.ac.uk>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: Re: xtreg v areg
Date   Mon, 14 Jun 2004 22:13:34 +0100 (BST)

Kit Baum wrote:

> Apples and oranges make a good fruit salad, but...
>
> The areg shows 11 region categories. The xtreg shows 304 groups of
> 'pano'. If you use the same variable (e.g. region) in both commands,
> you will get the same coefficients/std errors, or your computer is
> broken.

As a gangster from the East End of London might say, I've sorted it. I
used my PANO (constituency) variable as the fixed effect in both models,
after Scott Merryman's corrective suggestion. This got me the identical
results I should have had for both LSDV and FE. I could have used REGION
as my fixed effect, but the codes in this variable represent much larger
areas of the UK. I prefer the former, since one cannot get much more
precise regional FEs than parliamentary constituencies (unless you use
local government wards, but that would be masochistic)!

However, I noticed something in my two 'identical' models:

. areg edconch edround2 edround3 edtime edtimesq edyear edpollch lagconch
laglabch lagldmch clmargin ldmargin conplace edenp class if edmarker==1,
absorb(pano)

Number of obs =    1632
F( 14,  1314) =   69.42
Prob > F      =  0.0000
R-squared     =  0.7324
Adj R-squared =  0.6678
Root MSE      =  6.1165

[...]

. tsset pano edyear

. xtreg edconch edround2 edround3 edtime edtimesq edyear edpollch lagconch
laglabch lagldmch clmargin ldmargin conplace edenp class if edmarker==1,
fe

Fixed-effects (within) regression             Number of obs      =      1632
Group variable (i): pano                      Number of groups   =       304

R-sq:  within  = 0.4252                       Obs per group: min =         1
       between = 0.4157                                      avg =       5.4
       overall = 0.3644                                      max =        10

                                              F(14,1314)         =     69.42
corr(u_i, Xb)  = -0.5123                      Prob > F           =    0.0000

[...]

Now, I have fit the same models here (with the same FE in both) whilst
'switching off' any weighting and cluster options in both of them. How is
it that the R^2 is not the same in both models? Doubtless I'm overlooking
something. Ta.

CLIVE NICHOLAS        |t: 0(044)191 222 5969
Politics              |e: clive.nicholas@ncl.ac.uk
Newcastle University  |http://www.ncl.ac.uk/geps
*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index