Dear Statalisters,

I've two questions about the weighting matrix in two-step GMM. Theoretically the number of instruments relative to the groups, when big, will cause the non-singularity of the weighting matrix. The first question is: Why even diminishing the number of instruments the stata's warning about the singularity continues to appear? and the second is: Is it better in a case like that to utilise the one-step procedure or the Windmeijer correction is suficcient?

Thanks.