[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

AW: st: AW: constant in the suest command

From   "Martin Weiss" <>
To   <>
Subject   AW: st: AW: constant in the suest command
Date   Thu, 1 Oct 2009 14:49:03 +0200


This is a peculiarity of -suest- in connection with -regress-, described in
[R], p. 1803. But this constant is not the one that you are worried about.
As you can see from the -test- output in the last row, it is not part of its
output, even if you add option "constant" to -test-...

sysuse auto, clear

reg price weight length  /* 
*/ trunk turn, nocon
est store one

reg weight length  /* 
*/ headroom turn, nocon
est store two

suest one two

te [one_mean =two_mean ], common
te [one_mean =two_mean ], common constant


-----Ursprüngliche Nachricht-----
[] Im Auftrag von Flora Hofmann
Gesendet: Donnerstag, 1. Oktober 2009 14:44
Betreff: Re: st: AW: constant in the suest command

I am sorry if my question was too confused to answer. The main reason for
estimating the equations as a system is that I want to test cross equation

The problem is that I want to estimate the equations without a constant.
Here is what I did: 

qui reg lny1 b1x1 b2x2 b3x3 year_dummies country_dummies, noc
est sto y1

qui reg lny2 b4x4 b5x5 b6x6 year_dummies country_dummies, noc
est sto y2

qui reg lny3 b7x7 b8x8 b9x9 year_dummies country_dummies, noc
est sto y3

suest y1 y2 y3

The output however always reports a constant for each of the three equations
as well. For example under the heading 'y1_mean' it reports the coefficients
and s.e. for all included variables and then under the heading 'y1_lnvar'a
constant is reported. It is this constant that is puzzeling. I am probably
just missing some basic feature of -suest-. However, I couldn't find any
explanation for the reason and interpretation of this constant. 

flora hofmann

> Careful! The association between -suest- and -sureg- is rather loose, in
> spite of similar names. 
> You may want to show your output with the "constant" issue to enable
> Statalisters to answer your question...
> Martin
> Dear statalist,
> I want to estimate a system of equations. Since heteroscedasticity seems
> to
> be an issue and according to the 'Breusch-Pagan test of independence'
> there
> are no efficiency gains from -sureg- I decided to use -suest-.
> The equations are estimated by fixed effects. In order to be able to use
> -suest- I estimated the equations using -regress- and including a full set
> of time and country dummy variables while dropping the constant. However,
> -suest- keeps returning a constant. What constant is that? And how can it
> be
> interpreted?
> Thanks for your consideration (of this rather basic question)
> Flora Hofmann
> -- 

GRATIS für alle GMX-Mitglieder: Die maxdome Movie-FLAT!
Jetzt freischalten unter
*   For searches and help try:

*   For searches and help try:

© Copyright 1996–2015 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index