Estimating Differential Equation Models On Time Series

This paper explores the statistical properties of estimates proposed by Doreian and Hummon (1976) for systems of simultaneous differential equation systems. Since these procedures are iterative and computationally burdensome, alternative simpler methods are also explored via Monte Carlo Simulations. The simpler methods provide reasonably good diagnostics as to whether or not the more com plicated procedures are worthwhile. Problems of autocorrelation, interequation linkage, and error variance are systematically varied, and the effects of these varia tions on the parameter estimates are detailed.