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

st: RE: xi regress: noconstant option

From   "Nick Cox" <>
To   <>
Subject   st: RE: xi regress: noconstant option
Date   Mon, 14 Feb 2005 15:45:08 -0000

does nothing different from 

regress return country 

so perhaps you mean 

xi : regress return 

The latter model just fits 
the means of return for each country. 
It is essentially an alternative to 

Adding the -nocons- option just 
forces the mean for the first country, 
or more generally the category omitted, 
to 0 and leaves the other predictions 
unchanged. I can't think of a scientific
or statistical reason to do that, but 
that may be failure of imagination 
or just plain ignorance on my part. 

More generally, the rationale for
forcing through the origin is usually 
something like a law of physics 
which tells you that the relationship 
really should start there. Doesn't 
sound like your case. 


Yvonne Capstick
> I have two categories: "return" and "country", where return 
> is "numeric" and 
> "country" is a three letter code corresponding to the 
> country. I know that I 
> can use the "xi" function to create dummy variables and run 
> the regressions 
> all in one step, i.e.
> xi : regress return country
> To avoid the dummy variable trap, it automatically omits the 
> first group of 
> "country". I know that I can use the "char" function to 
> specify omitting 
> another group, e.g.
> char country[omit] "ita"
> would omit the group "ita".
> However, I haven't found a way to omit no groups - i.e. 
> include all the 
> dummy variables, and avoid the dummy variable trap by specifying 
> "noconstant" in the regression. If I try:
> xi: regress return country, noconstant
> it will omit the constant but still omit the first group of 
> "country". Any 
> advice would be appreciated.

*   For searches and help try:

© Copyright 1996–2021 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index