Some questions about the GMM notebook.

[cytwill]

Hi, everyone

I notice that in the notebook of GMM, the error matrix suggests that every element of the matrix is a difference between the moment of each observation in the data and the moment of our estimation. However, I think the moments are always statistics built on the whole dataset in GMM, not each observation. For example, when we choose the mean as the moment, we calculate the mean from all the observations not a single observation (this also seems to be impossible), right?

[davefoote]

I think the error vector for a moment like mean would just be the differences in means, but if its a large amount of arbitrary bins we have to calculate the difference in households we put in each bin, with each bin and simulated bin value being the moment. Not sure though.

[cytwill]

I understand what you mean by using bins @davefoote, like what is in the problem set. But I think the percentage of observations bins are moments, these moments are also derived from the whole data. I think in GMM we generate the data moments (no matter what and how many moments) from the whole dataset, instead of each observation.

[cytwill]

I have made it clear, there are no mistakes on the notebook. Thanks