Dear Statalisters,

When running a GMM in a 2-step procedure we use the optimal weighting matrix to weight the moments and better the assymptotic variance of the gmm estimator. The weighting matrix must be inverted and its rank is equal to the number of groups N, so because of that the number of instruments (moments) must be less than N, otherwise the matrix cannot be inverted.

The problem is: Even reducing the number of instruments making them less than N, the command xtabond2 continues to informe that the weighting matrix is singular. Why this occur? Is it a problem with the command or what?

Please if someone can help me I will really appreciate. I searched hardly something that could clarify this point for me but I found nothing.