OLS - problems with estimation equation

[hank83]

Hey everyone,

I have a problem with my OLS-regression. I´m trying to make a regression that looks like this:

icrge ln_gdp bb infl sacw ethnf assas ethnf_assas m2_1 ssa easia td1 td2 td3 td4 td5 td6 engfrac
eurfrac latitude

Where the 3 last variables are supposed to be instruments of the dependent variable. I know
that it´s possible to do a 2SLS directly in eviews, but Im not sure how to enterpret the result,
+ I wanna know the coefficients for the instruments.
My problem is these:

- first, it says: "alpha series in specification - engfrac" @expand can convert to categorical dummies.

so I re-write:

icrge ln_gdp bb infl sacw ethnf assas ethnf_assas m2_1 ssa easia td1 td2 td3 td4 td5 td6
@expand(engfrac, eurfrac, latitude)

but now it says "near singular matrix"
I guess that means that it is problem with multicollinearity, but I don´t know how to solve it.
I have tried to remove variables but the problem persist. And what if I don´t want to remove variables
until I have seen the result?

And what exactly is an alpha series? engfrac is just another data-series. And the only dummies,
except for td1-6 are "ssa" and "easia".


Im very grateful for comments and help, Im really stuck on this one..

Cheers,
Hannes

[hank83]

hi again,

btw, Im using eviews 5 if that could matter.

Cheers,

Hannes

[startz]

Your use of @expand gives a complete set of dummies. You probably have either another complete set or a constant.

[hank83]

Thanks for your answer Startz.

But I would prefer not to use @expand, because I don´t want to make dummies out of the last 3 variables, and as you said - I already have sets of dummy variables. I cant find my constant, don´t know exactly what to look for in the equation. Do you know how to make the equation work without @expand?

cheers,
Hannes

[startz]

What's your variable engfrac supposed to be? If it's supposed to be numbers, you need to convert it to a numerical series. (Look at the help system for "@val")

[hank83]

Hi,

you are absolutely right, the variable engfrac is a numerical value but has an "abc"-sign. I have looked @val up in the help-pdf, but unsure how to use it.
Do I press genr in the workfile or in the serie engfrac, and what exactly do I write? I find the help-pdf a bit unclear on that, because I don´t want to write just A specific value, but use the values that is already in the series.

Cheers,
H

[hank83]

ah, found out how to do it now! thanks a lot for your help!! =)

/H