Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

RE: st: why do fixed-effects estimates differ?


From   jverkuilen <jverkuilen@gc.cuny.edu>
To   <statalist@hsphsun2.harvard.edu>
Subject   RE: st: why do fixed-effects estimates differ?
Date   Wed, 10 Sep 2008 12:10:33 -0400

I think Maarten's right, it's probably differences due to finite precision arithmetic. Try using doubles and see if it goes away. 

This kind of thing is hardly unusual and is even more marked in nonlinear problems such as numerical optimization, where you simply can't expect more than about eight digits in double precision for many problems. Root finding for the derivative of a function is inherently harder than for the function itself.   

-----Original Message-----
From: "G.B.Li" <tslee1912@yahoo.com>
To: statalist@hsphsun2.harvard.edu
Sent: 9/10/2008 3:27 AM
Subject: st: why do fixed-effects estimates differ?

Hi all:

I use two methods to control for fixed effects: (1) use dummy variables, (2) demean all the variables.

Results turn out to be highly close, but not identical. The difference is no more than 0.000001。

In theory, they should be identical. How to explain the trivial differences?

(I guess it is related to my huge sample size. When I try both methods with a very small dataset, there is no such difference.)

Many thanks!
GBL


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   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