multi- collinearity

[Afsaneh]

Hi,
I'm estimating a regression with 10 independent variables. When I run the regression I get a message "near singular matrix". I reviewed other posts on this message and I realized it shows multi collinearity and I need to check VIF test. But how can I do that? Because I need to estimate regression first and then run VIF test! However, as I said Eviews 8 prevent to estimate the equation with getting that message"near singular matrix". Also, i read in one post dropping the colineared variable is a big mistake, then after finding collineared variables how can I fix this problem? I would greatly appreciate any help.

[startz]

The odds are very strong that you've made a mistake in specifying the regression. You might post your EViews workfile and the exact command you use to see if anyone can spot something.

[Afsaneh]

I attached the workfile. I simply used this command: r c bm dy ep inf svar pres ms gby tb ts , r as a dependant and 10 others as independent. This command works if I eliminate gby that I guess, it might be correlated with 2 others.

[startz]

Yeah, GBY=TB+TS. So GBY is redundant.

[Afsaneh]

But how could I do FVI test before deleting gby?

[startz]

I guess you can't. But why would you want to? The near perfect multicollinearity message means you have a logically redundant variable (with very rare exception).