Page 1 of 1
variance inflation factor
Posted: Fri Aug 09, 2013 10:48 am
by lnp3
dear forum,
I am using Eviews 8. How can I test multi-collinearity? I am always using pairwise correlation matrix to control MC (among independent variables) and dropping variables that cause high inter-correlation (let's say over 0.4; 0.5)
however, Is there any test of VIF? variance inflation factor? If so, what is the determination criterion?
as always, I highly appreciate your help.
lnp3
Re: variance inflation factor
Posted: Fri Aug 09, 2013 10:55 am
by EViews Gareth
After you've estimated your equation, you can click View->Coefficient Diagnostics->Vif
Re: variance inflation factor
Posted: Fri Aug 09, 2013 1:09 pm
by lnp3
thank you Eviews Gareth.
Re: variance inflation factor
Posted: Sun Aug 11, 2013 4:35 am
by lnp3
dear forum,
how can i read these variance inflation factors in Eviews 8? when are variables considered to be multi-collinear?
Variance Inflation Factors
Date: 08/11/13 Time: 07:32
Sample: 2006 2012
Included observations: 178
Coefficient Uncentered
Variable Variance VIF
D(TNINTEX) 0.062030 2.016873
D(MS) 4.97E+13 2.446379
D(RIGL) 3.63E+08 1.417163
D(TCGDP) 1.21E+09 2.824274
D(MBGDP) 3.27E+10 1.950917
D(FFGDP) 5.90E+08 1.339142
thanks a lot in advance.
Re: variance inflation factor
Posted: Sun Aug 11, 2013 8:10 am
by startz
Why do you care if the variables are multi-collinear?
Re: variance inflation factor
Posted: Mon Aug 12, 2013 9:05 am
by lnp3
it is asked in a review report.
Re: variance inflation factor
Posted: Mon Aug 12, 2013 9:28 am
by startz
lnp3 wrote:it is asked in a review report.
Hard to argue with that
Re: variance inflation factor
Posted: Mon Aug 12, 2013 9:33 am
by lnp3
I was informed that Gujarati says that VIF> 10 then Multicoloinarity exists,and some others restrict it to 5 however we take 10.
Is this rule for entire test? or individual variables? Eviews does not report the entire test. It reports for individual variables. Please see results above.
I hope somebody can help me with this?
thanks a lot.