Page 1 of 1
Panel data Coefficient Variance Decomposition Bug?
Posted: Mon Jul 10, 2017 1:38 am
by acemi
Hello,
Variance decomposition for eq02 is attached within workfile as table fi_coefvardecomp.
Somehow eviews displays coefficients (c1 to c5) rather than regressors themselves in the table.
How are we to determine multicoll between regressors with this table please comment.
Thank you
Re: Panel data Coefficient Variance Decomposition Bug?
Posted: Mon Jul 10, 2017 7:47 am
by EViews Gareth
Your equation was specified by coefficient, not by regressors.
Re: Panel data Coefficient Variance Decomposition Bug?
Posted: Mon Jul 10, 2017 10:51 am
by acemi
So how can i interpret the results any idea.
Re: Panel data Coefficient Variance Decomposition Bug?
Posted: Mon Jul 10, 2017 1:38 pm
by EViews Gareth
Match up C(1) with the C(1) in the equation output.
Re: Panel data Coefficient Variance Decomposition Bug?
Posted: Mon Jul 10, 2017 9:20 pm
by acemi
OK I matched it and then does it give me info on multicorrelation of c(1)'s regressor with other coefficient's regressors for example ?
Re: Panel data Coefficient Variance Decomposition Bug?
Posted: Mon Jul 10, 2017 9:34 pm
by EViews Gareth
Yes.
Re: Panel data Coefficient Variance Decomposition Bug?
Posted: Mon Jul 10, 2017 9:36 pm
by acemi
I have one issue : as you can see from attached file eq02 on previous posts, the regression specification has coefficients (c(1) to c(5)) jointly multiplying the regressors. For example : (1-exp(c(1))@trend)*c(2)*(1-c(3))*LOG(Saving_rate*100)
Do you have any idea how are we to interpret the Variance Decomposition results within this setting ?
Thank you.