Smoothing time-series data by nonmetric polytone curves

Abstract

This paper suggests a simple nonmetric method for smoothing time series data. The smoothed series is the closest polytone curve to the presmoothed series in terms of least sum of absolute deviations. The method is exemplified on several seasonally adjusted series in order to estimate their trend component.

Keywords:

• Monotone Curve
• Polytone Curve
• Seasonally-adjusted data
• Trend component