Empirical comparison of finite sample distribution of the least squares estimator in a non-statonary autoregressive model

Abstract

This paper examines the finite sample distribution of the least squares estimator in 3 first order nonstationary autoregressive model. The finite sample distributions are computed using a convenient algorithm developed by Tsui and Ali (1992a, b). The exact distributions are compared with the Satchell's (1984) Edgeworth-type approximations, the Pearson distributions and the limiting distributions of the least squares estimator. The Pearson distributions are found to be more adequate than Satchell's approximations.

Keywords:

• Characteristic function
• Satchell's Edgeworth-type approximation
• Pearson distribution
• Cauchy distribution