Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

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

RE: st: obtaining R squared after xtabond

From   Nick Cox <>
To   "''" <>
Subject   RE: st: obtaining R squared after xtabond
Date   Wed, 2 Feb 2011 16:51:57 +0000

No doubt you can get R-squared to more decimal places, for what it's worth. 

My main answer is the same: use -test- if possible. 


Anastasiya Zavyalova

Hey Nick,

Thank you for the quick response.

I have calculated R sq for both of my models after the xtabond  
estimation. They are  both 0.99, but Model 2 has five more variables  
than Model 1.

For Model 1: Wald chi2(21)         =   3664.36  Prob > chi2            
=    0.0000
For Model 2: Wald chi2(26)         =   3867.88  Prob > chi2            
=    0.0000

How can I find out from this information how much more variance is  
explained by Model 2 than Model 1 and which model has the best fit? Is  
there a way to compare whether the two Chi sqrd statistics are  
significantly different?

On Feb 2, 2011, at 4:27 AM, Nick Cox wrote:

> It's best not to try assessing significance from R-squared. You should
> try to test the extra variables for significance directly. -test- I
> imagine to be the way to do it.
> Otherwise, the difference in R-squared is just that. Calculate both
> and subtract. I don't see what your difficulty is there.  The recipe
> is not
> correlate and square predicted and actual dependent variable
> but
> correlate predicted and actual dependent variable, and square
> Nick
> On Wed, Feb 2, 2011 at 2:05 AM, Anastasiya Zavyalova
> <> wrote:
>> Hey all,
>> How can I obtain an R-squared statistic after I run xtabond? I did  
>> the old
>> school: correlated and squared predicted and actual dependent  
>> variable. The
>> only problem is that I have two models, where Model 2 contains all  
>> the
>> variables form Model 1 plus a couple more.  Reviewers need to know  
>> how much
>> more variance Model 2 explains over Model 1. So: 1) how can I find  
>> out how
>> much variance each model explains and 2) whether Model 2 explains
>> significantly more variance than Model 1?

*   For searches and help try:

© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index