Algorithm for Positive Definite Least Square Estimation of Parameters

摘要

An algorithm for positive definite least square estimation of parameters is presented. This estimation problem arises from the PILOT dynamic macro-economic model and is equivalent to an infinite convex quadratic program. It differs from ordinary least square estimations in that the fitting matrix is required to be positive definite. The algorithm solves the infinite convex quadratic program by generating and solving a sequence of ordinary convex quadratic programs. By specifying a constant, the algorithm will find an approximate optimal solution after finitely many iterations, or will tend to an optimal solution in the limit. The algorithm is generalized to solve a class of infinite convex programs. 4 refs., 2 tabs. (ERA citation 11:037247)