Abstract
Koreisha and Pukkila (1990a) have recently proposed a computationally convenient three-step GLS-type linear estimator for the regression model with ARMA disturbances involving three sequential applications of least squares. One potential drawback to this estimation procedure is that it entails dropping a significant number of initial observations. This paper uses Monte Carlo methods to evaluate its performance vis-à-vis existing OLS and GLS linear estimators.
Original language | English (US) |
---|---|
Pages (from-to) | 315-332 |
Number of pages | 18 |
Journal | Communications in Statistics Part B: Simulation and Computation |
Volume | 26 |
Issue number | 1 |
DOIs | |
State | Published - Jan 1 1997 |
Keywords
- Approximate GLS estimator
- Autocorrelation
- Regression
ASJC Scopus subject areas
- Statistics and Probability
- Modeling and Simulation