Class LoessFilter

  extended by peakml.math.filter.LoessFilter
All Implemented Interfaces:

public class LoessFilter
extends java.lang.Object
implements Filter

Implementation of a LOESS filter (also known as local regression) for smoothing signals. The method is also called Locally Weighted Polynomial Regression, because at each point in the dataset a low-degree polynomial is fit to a subset of the given window-size. The polynomial function is fitted to the data using weighted least squaresgiving more weight to points near the point whose response is being estimated and less weight to points further away. The value of the regression function for the point is then obtained by evaluating the local polynomial using the explanatory variable values for that data point. The LOESS fit is complete after regression function values have been computed for each of the n data points.

The disadvantage of the LOESS filter is that it requires fairly large, densely sampled datasets in order to produce good models. Furthermore, the method is known to be fairly computational intensive.

Constructor Summary
LoessFilter(double windowsize)
          Standard constructor, which accepts the window-size to use for the smoothing filter.
Method Summary
 double[] filter(double[] xvals, double[] yvals)
          With this method actual filtering is performed.
Methods inherited from class java.lang.Object
equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail


public LoessFilter(double windowsize)
Standard constructor, which accepts the window-size to use for the smoothing filter.

windowsize - The windowsize expressed in percentage of the number of data-points.
Method Detail


public double[] filter(double[] xvals,
                       double[] yvals)
                throws java.lang.IllegalArgumentException
Description copied from interface: Filter
With this method actual filtering is performed. The xvals and yvals arrays should be of equal length, otherwise an IllegalArgumentException is thrown. The return value is the new yvals array, with smoothed out values. The returned array is of the same size as the x and y value arrays.

Specified by:
filter in interface Filter
xvals - The x-values of the signal.
yvals - The y-values of the signal.
The resulting smoothed version of the y-values.