A simplified approach to M-estimation with application to two-stage estimators (Q1088322)

An M-estimator \({\hat \theta}\) is one obtained by solving the system of equations \(\sum^{n}_{i=1}\Psi (z_ i;\theta)/n=0\), where \(z_ i\) are the data vectors and \(\theta\) an unknown parameter vector. The ordinary least squares estimators, instrumental variables estimators and the usual sort of maximum likelihood estimators with interior solutions are all M-estimators. This paper presents a simple version of the theory of M-estimators.