R-squared

For econometric discussions not necessarily related to EViews.

Moderators: EViews Gareth, EViews Moderator

selin1000
Posts: 40
Joined: Tue Mar 06, 2012 1:56 pm

R-squared

Postby selin1000 » Fri Aug 03, 2012 1:05 pm

Hi everybody!

I just have a quick question. I have been running cross-sectional OLS regressions, and my coefficients are significant (p-value almost zero) however I realize that I often get strange R-squared values which are either very close to zero or negative. Is this a problem with Eviews or does that mean that my regressions do not work?

startz
Non-normality and collinearity are NOT problems!
Posts: 3775
Joined: Wed Sep 17, 2008 2:25 pm

Re: R-squared

Postby startz » Fri Aug 03, 2012 1:08 pm

Have you left a constant out of the regression?

selin1000
Posts: 40
Joined: Tue Mar 06, 2012 1:56 pm

Re: R-squared

Postby selin1000 » Fri Aug 03, 2012 1:09 pm

Yes I have.

selin1000
Posts: 40
Joined: Tue Mar 06, 2012 1:56 pm

Re: R-squared

Postby selin1000 » Fri Aug 03, 2012 1:10 pm

Now I tried with a constant but my R-squared value is 0.000352 :(

startz
Non-normality and collinearity are NOT problems!
Posts: 3775
Joined: Wed Sep 17, 2008 2:25 pm

Re: R-squared

Postby startz » Fri Aug 03, 2012 1:12 pm

There is not necessarily anything wrong, but you might post a screen shot of your output, along with describing what your variables are, to see if anyone has any suggestions.

selin1000
Posts: 40
Joined: Tue Mar 06, 2012 1:56 pm

Re: R-squared

Postby selin1000 » Fri Aug 03, 2012 1:17 pm

Thank you very much :) So I paste one of my regressions. I am just regressing a portfolio's total beta on its return as below. (total portfolio beta=betam1+betam2-betam3-betam4 and RET=portfolio return)
I'm happy about the positive and significant coefficient but am just worried about the strange R-squared value. Thanks a lot again!

Dependent Variable: RET?
Method: Pooled Least Squares
Date: 08/03/12 Time: 23:11
Sample: 11 1910
Included observations: 1900
Cross-sections included: 8
Total pool (balanced) observations: 15200

Variable Coefficient Std. Error t-Statistic Prob.

BETAM1?+BETAM2?-BETAM3?-BETAM4? 0.034173 0.012899 2.649198 0.0081

R-squared 0.000345 Mean dependent var 4.81E+09
Adjusted R-squared 0.000345 S.D. dependent var 4.44E+11
S.E. of regression 4.44E+11 Akaike info criterion 56.47765
Sum squared resid 3.00E+27 Schwarz criterion 56.47815
Log likelihood -429229.1 Hannan-Quinn criter. 56.47781
Durbin-Watson stat 2.001845

startz
Non-normality and collinearity are NOT problems!
Posts: 3775
Joined: Wed Sep 17, 2008 2:25 pm

Re: R-squared

Postby startz » Fri Aug 03, 2012 1:21 pm

You need to include a constant for this to make sense. If you are still getting a signicant coefficient with a constant included, then everything's probably okay. You have a very large number of observations which is what leads to a low R-square with significant t-stats.

selin1000
Posts: 40
Joined: Tue Mar 06, 2012 1:56 pm

Re: R-squared

Postby selin1000 » Fri Aug 03, 2012 2:26 pm

Now I understand :) Thank you so much for the clarification! Have a wonderful day!

Selin


Return to “Econometric Discussions”

Who is online

Users browsing this forum: No registered users and 22 guests