An improved version of the Koopmans-levin algorithm

摘要

In this paper two algorithms developed by Koopmans and Levin are discussed in details. One of them eliminates unwanted correlation in the input data by dropping some I/O pairs, while the other has the capability to completely utilize all the available input/output records. As a main result a new estimation algorithm is derived in this paper leading to maximum likelihood estimation and simultaneously utilizing the available input/output records with maximum efficiency. Simulation examples verify the exceptionally good performance of the new algorithm.