Error: estimated covariance matrix of moment conditions not of full rank, and optimal GMM weighting matrix not unique.

[Nicolas Vamvas]

Hello,

I am trying to run the following regression:

by id: ivreg2 a (b c d=iv*), gmm2s robust small first ffirst

The problem is that I always get an error for id = 15, namely
"Error: estimated covariance matrix of moment conditions not of full rank,
and optimal GMM weighting matrix not unique.
Possible causes:
singleton dummy variable (dummy with one 1 and N-1 0s or vice versa)
partial option may address problem."

I don't use any dummy variable.

Could someone help me to solve this problem please?

Thank you.

[Mark Schaffer]

You need to look at your data and try to deduce why you are not getting a full-rank matrix. One thing you could try is estimating by standard IV (so drop the gmm2s option), save the long-run covariance matrix, try inverting it, and then look at it:

mat S=e(S)
mat Sinv=invsym(S)
mat list S

If there's a row or column of zeros, it could identify the problematic variable.