Omitted because of collinearity in regression

[Mirja Kvarnevik]

Hi!

I'm working with data from the Eurobarometer, and the data set has a variable for regions of the eu (cont, south, north, east).
I've used the variables in regressions, and all has been fine, but now when I need to re-run the regression I get the collinearity error for one of the region variables. The weird thing is that it happens to which ever variable that is last in the regression command. When I just use three of them, the error doesn’t occur.

When I use the variable that was "omitted" as the dependant variable, the other regions comes out as -1. But the other variable comes out as zero (more or less).

Does anyone out there in the statverse have an idea about what's going on? And how I can get it right?
Would be very helpful.
Thanks!

[Fei Wang]

My bold guess is that the four variables (cont, south, north, east) are perfectly collinear and always sum up to one. You have to drop one category in this case.

[Carlo Lazzaro]

Mirja:
I do share Fei's idea.
More helpful replies are conditional on posting what you typed and what Stata gave you back, along with an example/excerp of your dataset that you can share via -dataex-.