Please click here if you are unable to view this page.
TOPIC:
ROBUST AND EFFICIENT SEQUENTIAL GEL AND GMM ESTIMATION AND TESTS WITH MARTINGALE DIFFERENCES
ABSTRACT
This paper considers sequential generalized empirical likelihood (GEL) with martingale differences and compare it with a corresponding sequential generalized method of moments (GMM) estimation and tests when there is a computationally simple pn-consistent estimator of nuisance parameters or the nuisance parameters can be eliminated with an estimating function of parameters of interest. As an initial estimate might have asymptotic impact on final estimates, we propose a general C(α)-type formulation to eliminate the impact. This formulation can be used in both the GEL and GMM frameworks to construct estimation and tests robust to initial estimates. The sequential approaches can save computational burden as the numbers of moments and parameters are reduced. The properly constructed robust sequential GEL (RSGEL) and robust sequential GMM (RSGMM) estimators of parameters of interest can be as efficient as corresponding joint GMM estimators. The RSGEL removes several higher order bias terms of the RSGMM. Our moment conditions at the true parameters are martingales, thus they cover some spatial and time series models. We investigate tests for parameter restrictions in RSGEL and RSGMM frameworks. Tests in the sequential frameworks can be locally as powerful as those in the joint moments estimation frameworks when the sequential estimators are efficient.
Keywords: Sequential estimation, GMM, empirical likelihood, martingale differences, high order bias, efficiency, test.