EViews.com

Hi everyone, i need some help as i don't know how to check if there is multicollinearity in the data. i used the logarithm of imports as the dependent variable and the logarithms of GDP and CPI only as explanatory variables. then i ran the multi regression on eview and i got these data, from these data, i want to know how to check if there is multicollinearity in the data and how to calculate the correlation matrix of the variables, anyone knows how to do that? here is the url file to the data i got from eview

http://www.mypicx.com/10252010/multi/

You should refer to your textbook for the discussion of collinearity issue. As for EViews, it offers you two very useful diagnostic tools: 1) Variance Inflation Factors and 2) Coefficient Variance Decomposition.

Out of curiousity, why do you want to check for multicollinearity?

Btw, are you using GDP to model Imports? Isn't the latter part of the former? I am not an economist anymore but I think that might be a big problem if you go ahead with this.

Ronny

this is how you do it:

go to Quick-> Group statistics -> correlations... then choose the independent variables you want to check i.e cpi and gdp.

you will get a correltion matrix. at the diagonal you will se all aces "1". you ignor these number since they represent the correlation of each independent variables with itself, so they are 1 by default. now you check the rest cells of the matrix. if you see correlations figures baove 0.7 or below -0.7 then you have problem , thus multicollinearity. in general the correlation is form -1 to +1 . the limit of 0.7 is typical in timseries analysis. if multicolinearity exists then you omit one of the variables with high correlation.

I hope tha above is of assistance.

Bill

Ronny

this is how you do it:

go to Quick-> Group statistics -> correlations... then choose the independent variables you want to check i.e cpi and gdp.

you will get a correltion matrix. at the diagonal you will se all aces "1". you ignor these number since they represent the correlation of each independent variables with itself, so they are 1 by default. now you check the rest cells of the matrix. if you see correlations figures baove 0.7 or below -0.7 then you have problem , thus multicollinearity. in general the correlation is form -1 to +1 . the limit of 0.7 is typical in timseries analysis. if multicolinearity exists then you omit one of the variables with high correlation.

I hope tha above is of assistance.

Bill

This is terrible advice. First, if you must check for multicollinearity use the EViews tools that Trubador describes. Second, there is no reason why 0.7 is an appropriate number. Third, while there are only two variables here, in the general case of more than 2 RHS variables looking at simple correlations is not an appropriate check for multicollinearity.

And most importantly, dropping a variable is usually a serious mistake.

I am using Eviews 11 student version lite. I want to test multicollinearity through VIF but I am unable to find the option.Can anyone please help me how to do VIF test

What have you done so far?

what I have done is attaching here

VIFs are not available for panel data.

Then how can I check multicollinearity ,can you please help me. I have some paper where students did VIFs for panel data. here I am giving what I have got from Eviews help.and can you please tell me that I want to check multicollinearity for fixed effect model and is the follwiing correlation matrix is fine?

EViews has no multicolinearity measures built in for panel data

Just out of curiosity, why do you care if there is multicollinearity here?