milan lee posted on Friday, September 12, 2014 - 9:15 am
Dear all, What's the rationale for using observed and latent covariates in GMM, respectively? It seems that observed covariates are much more often used in current GMM publications. If both options are fine, how to tell which one is better? I have now two covariates (xm, xn) and each was aggregated by two observed variables (xm1 & xm2, xn1 & xn2). I prefer to use observed covariates but because I am new to this approach, I don't know how to justify it if people ask why latent covariates are not used.