Library Header Image
LSE Theses Online London School of Economics web site

Comparative methods of computing maximum likelihood estimates for non-linear econometric systems

Chong, Yock Yoon (1981) Comparative methods of computing maximum likelihood estimates for non-linear econometric systems. PhD thesis, London School of Economics and Political Science.

[img] Text
Download (6MB)


This research is mainly concerned with numerical optimisation techniques applied to general non-linear econometric simultaneous equations systems. The method of estimation used is maximum likelihood. An estimation program which applies gradient-type procedures, specifically the Berndt-Hall-Hall-Hausman and Gill-Murray' Pitfield methods, is developed. This program allows the estimation of a general small-to-medium size model which is non-linear in parameters, variables or both. In the course of program development, a general differentiation program is written which will differentiate a set of econometric equations and thus provide the analytical gradients for the optimisation procedures. A comparative study has been made of the relative efficiency of the two methods by running a set of simulated non-linear models and also using a small macro- economic model of the British Economy specified by David F. Hendry. To improve the efficiency of the estimation program in terms of computing time, the Berndt-Hall-Hall-Hausman method was implemented on the ICL Distributed Array Processor (DAP)’ which employs parallel computations. The DAP runs show that for a model with a large sample size, the DAP is approximately 30 times faster than the conventional computer CDC 7600, but that for the present algorithm, the latter is a more efficient alternative for small sample sizes.

Item Type: Thesis (PhD)
Additional Information: © 1981 Yock Yoon Chong
Library of Congress subject classification: H Social Sciences > HB Economic Theory
Q Science > QA Mathematics

Actions (login required)

Record administration - authorised staff only Record administration - authorised staff only


Downloads per month over past year

View more statistics