Australian Journal of Statistics (1997), 39, 313-325.
Rob J Hyndman1 and Matt Wand2
- Department of Econometrics and Business Statistics, Monash University, Clayton VIC 3168, Australia.
- Harvard School of Public Health, Harvard University, Massachusetts, USA.
Abstract: Nonparametric estimators of autocovariance functions for non-stationary time series are developed. The estimators are based on straightforward nonparametric mean function estimation ideas and allow use of any linear smoother (e.g. smoothing spline, local polynomial). We study the properties of the estimators and illustrate their usefulness through application to some meteorological and seismic time series.
Keywords: bandwidth; correlated errors; kernel smoothing; local polynomial; nonparametric regression; non-stationary model; time series.