LM Test with 0 R^2 Coefficient and Probabilities=1

[JTrudeau]

Hi,

I was having no trouble with my Eviews until recently when I went to conduct a series of LM tests that when I tried to regress the equation against the residual as the dependent variable I got 0.0000 for the R^2 coefficient and 1.0000 for all of the probability values. I've tried this in multiple data sets now, and don't know what happened to prevent me from getting answers as I had before. Any help is much appreciated!

-Jenn

Example:


Dependent Variable: U3
Method: Least Squares
Date: 04/25/09 Time: 11:15
Sample: 1 51
Included observations: 51

Variable Coefficient Std. Error t-Statistic Prob.

C 7.28E-11 6.234733 1.17E-11 1.0000
LAGED85 -8.88E-12 13.78864 -6.44E-13 1.0000
LSSIPYMT -8.53E-12 0.728014 -1.17E-11 1.0000
W -7.81E-14 0.112336 -6.96E-13 1.0000

R-squared 0.000000 Mean dependent var -2.17E-15
Adjusted R-squared -0.063830 S.D. dependent var 0.360109
S.E. of regression 0.371424 Akaike info criterion 0.932238
Sum squared resid 6.483912 Schwarz criterion 1.083754
Log likelihood -19.77207 Hannan-Quinn criter. 0.990137
F-statistic 8.58E-15 Durbin-Watson stat 1.774349
Prob(F-statistic) 1.000000

[startz]

Hi,

I was having no trouble with my Eviews until recently when I went to conduct a series of LM tests that when I tried to regress the equation against the residual as the dependent variable I got 0.0000 for the R^2 coefficient and 1.0000 for all of the probability values. I've tried this in multiple data sets now, and don't know what happened to prevent me from getting answers as I had before. Any help is much appreciated!

-Jenn

Example:


Dependent Variable: U3
Method: Least Squares
Date: 04/25/09 Time: 11:15
Sample: 1 51
Included observations: 51

Variable Coefficient Std. Error t-Statistic Prob.

C 7.28E-11 6.234733 1.17E-11 1.0000
LAGED85 -8.88E-12 13.78864 -6.44E-13 1.0000
LSSIPYMT -8.53E-12 0.728014 -1.17E-11 1.0000
W -7.81E-14 0.112336 -6.96E-13 1.0000

R-squared 0.000000 Mean dependent var -2.17E-15
Adjusted R-squared -0.063830 S.D. dependent var 0.360109
S.E. of regression 0.371424 Akaike info criterion 0.932238
Sum squared resid 6.483912 Schwarz criterion 1.083754
Log likelihood -19.77207 Hannan-Quinn criter. 0.990137
F-statistic 8.58E-15 Durbin-Watson stat 1.774349
Prob(F-statistic) 1.000000

I may misunderstand the question, but if you regress residuals on the right hand side variables from the equation that generated the residuals, you should get an R-square of zero. The residuals are uncorrelated with the RHS variables.

[JTrudeau]

I'm pretty sure that the standard method to finding the LM stat is to run a regression
LS Y C X1 X2 X3...
Then generate a variable for the residuals
GENR U = RESID
Then regress the equation with the residual term as the dependent variable
LS U C X1 X2 X3 ...

Am I mistaken?

[EViews Gareth]

Startz is right, if you regress the residuals on the regressors, you will get an R-squared of zero (and all the other strange results you're seeing) by construction.

Obviously we can't tell you what the form of your LM Test is supposed to be, since we don't know what you're testing, but it is highly unlikely that you're supposed to be regressing the standard residuals on all of the original regressors.

[w2quesang]

Hi just wanna ask if what I am doing is correct.

In the version 4 of the Eviews, to test the ARCH-LM of a time series data, I just simply type "r c" on the LS-least squares estimation then test the residuals for its ARCH-LM effect? r is the name of the data series.

However, in the version 5 of Eviews, it seems I cant do the same because there is no test for ARCH-LM effect on the same procedure.

Just wanna ask about the approach of you guys on this. Basically I needed to check whether my time series data shows an ARCH effect in order to justify using a GARCH for it.

Thanks

P.S. What will also be my indicator to know that the data has or didn't have ARCH effect. Thanks

[sdi]

pls, i need help! some1 analyzed a data using daily entries, tested for arch effect and concluded that arch errors are present, but when i did d same with monthly data, same company, i got an response of "no arch errors". The procedure i used is as follows:
1. i estimated equation y = c + r :y is dependent var, c is constant, r is residual
2. then i did view/residual test/arch lm test.
I USE EVIEWS 3.0
WHERE DID I GO WRONG? THE GRAPH SUGGESTS THAT ARCH ERRORS EXISTS. PLEASE HEEEEEEEEEEEEEEEEEEEEEELP!