A new linear phase, robust, non-linear smoothing method for stationary and non-stationary time or space series data

A description is given of a linear-phase robust nonlinear smoothing method (the LRNS method) that can be used to smooth any time-series or space-series data to any degree. After a short review of existing smoothing methods, the drift removal method is discussed, its limitations are shown, and a step-by-step description is given how they are overcome in the proposed LRNS method. Examples are used to demonstrate the effectiveness of this method in smoothing synthetic and real data. With the LRNS method, the degree of smoothing is controlled by choosing only one parameter; any data series can be smoothed, whether it is regularly or irregularly sampled, contains spikes, random noise or missing data; only one weight is used this weight automatically adapts with changes in the data. Efficiency and stability are unaffected by the degree of smoothing, and no amplitude overshooting occurs. The method's greatest benefit lies in its simplicity rival those of 'smoothing by eye' without its bias and subjectivity.<>