fixed effect error message "near singular matrix"

[Mya]

Hey, I'm estimating a gravity model (exports between countries), I have variables which vary over time and cross section (GDP, POP...) and I have constant variables distance between two countries for example. When I add these constant variable and I run fixed effect I get an error message "near singular matrix". how can I deal with that i can't delate these two variables they are important. Please need your help.

[startz]

You cannot have both fixed effects and variables which are constant in the cross-section. It's meaningless.

[Mya]

Thank you for your answer. It's the case in gravity model of exports between pairs of countries, the GDP, population... varies over time and cross section but the distance is constant over the time and there are dummy variables too which doesn't vary over time. I don't really know how to deal with that I can't drop them they are imprtant

[startz]

Either drop distance or drop the dummies. Those are the only choices.

[indah284]

Hi, guys.

Same with Mya, I'm estimating a gravity model too. It's about export in 4 cross sections over 10 years. I have variables which vary in time and cross section and I have constant variable (distance between two countries). I have read that I can't estimate with fixed effect because of constant variable, but later I try to estimate with random effect and it doesn't work. I get error message "Random effects estimation requires number of cross section>number of coefs for between estimator for estimate of RE innovation variace."

So what should I do with this? I hope you could help me because I have to finish my thesis.
Thank you.