Climate and Dynamics
Limitations on regression analysis due to serially correlated residuals: Application to climate reconstruction from proxies
Article first published online: 27 SEP 2005
Copyright 2005 by the American Geophysical Union.
Journal of Geophysical Research: Atmospheres (1984–2012)
Volume 110, Issue D18, 27 September 2005
How to Cite
2005), Limitations on regression analysis due to serially correlated residuals: Application to climate reconstruction from proxies, J. Geophys. Res., 110, D18103, doi:10.1029/2005JD005895., and (
- Issue published online: 27 SEP 2005
- Article first published online: 27 SEP 2005
- Manuscript Accepted: 1 JUL 2005
- Manuscript Revised: 4 MAY 2005
- Manuscript Received: 25 FEB 2005
- regression methods;
- climate reconstructions
 The effects of serially correlated residuals on the accuracy of linear regression are considered, and remedies are suggested. The Cochrane-Orcutt method specifically remedies the effects of serially correlated residuals and yields more accurate regression coefficients than does ordinary least squares. We illustrate the effects of serially correlated residuals, explain the application of the CO method, and evaluate the gains to be achieved in its use. We apply the method to an example from climate reconstruction, and we show that the effects of serial correlation in residuals are present and show the significantly improved result.