A0361
Title: New second-order asymptotic methods for nonlinear models
Authors: Paul Rilstone - York (Canada)
Gubhinder Kundhi - Memorial University of Newfoundland (Canada) [presenting]
Abstract: New higher-order asymptotic methods for nonlinear models are developed. These include generic methods for deriving stochastic expansions of arbitrary order, methods for evaluating the moments of polynomials of sample averages, a method for deriving the approximate moments of the stochastic expansions, simplified expressions for the first four moments of nonlinear estimators, a third-order approximate Saddlepoint expansion. The techniques are applied to improve inferences with the Two-Stage Least Squares (2SLS) estimator and the weak instruments problem. In a Monte Carlo simulation experiment, higher-order analytical corrections provided by the Saddlepoint approximation are compared to the Edgeworth, the bootstrap and the first-order approximation in finite samples in an i.i.d sampling context.
