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

Re: st: areg vs xtreg, fe

From   David Jacobs <>
Subject   Re: st: areg vs xtreg, fe
Date   Mon, 04 Feb 2008 13:33:50 -0500

The main reason for the difference is that AREG counts the machine equivalent of the case dummies when it computes the R2; EXTREG, FE does not. It follows that a lion's share of the variance you've explained is due to the case specific attributes you've included in your models.

Two incidental matters: First, if you end up using XTREG, FE the proper R2 to report is the "within". Second, R2 square values aren't the best measure of anything that matters although referees sometimes think they matter . The point is I wouldn't spend much space in your write up on these values.

Dave Jacobs

At 12:22 PM 2/4/2008, you wrote:

Dear all,

hi I write to ask you one question regarding the difference between areg and xtreg, fe. In particular I would like to understand the difference concerning the r squared. Indeed I use a fixed effects model to estimate the returns to education on a panel of individuals. I performed fixed effects estimates for two categories of workers, blue collars (which are many) and white collars.
I found out interesting results. However, for blue collars the estimates of the r-squared with xtreg,fe are the following:
r^2 within 0.07
r^2 between 0.002
r^2 overall 0.004
when I do the same estimation with areg, I obtain an r^2 of 0.80. a huge difference.

I would like to understand why is there a so big difference, and which one I should report...

Another related question concern the estimates for white collars. The xtreg, fe have an overall r^2 of 0.27 (much better than the previous one). However when I try the areg command it says that the matrix is not positive definite (I use also some lagged values of my variable of interests in the estimation).

Could anyone help me?
many thanks

* For searches and help try:

*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index