Lowess Nonparamatic Regression
Lowess Regression (LOcal Weighted regrESSion) does not require a mathematical function and does not require parameters (non-parametric). This fitting method is especially useful for data that does not conform to a specific mathematical model or the mathematical model is unknown.
Select Lowess Regression from the Function list.
After selecting the Lowess method simply press the fit button.
to perform the Lowess regression.
A Lowess regression section is added to the graph options scroll view. The user as the option to adjust the extent of Lowess smoothing, which is a fraction between 0 and 1.
F: Specifies the amount of smoothing, as F increases the curve becomes smoother. F values between 0.20 to 0.75 usually result in a good fit.
After the initial Lowess fit, the F parameter can be adjusted simply by moving the slider control.
For more information on Lowess Regression see Cleveland, W. S. (1979), “Robust locally-weighted regression and smoothing scatterplots,” Journal of the American Statistical Association, 74, 829-836
The Loess regression is a extension of the Lowess method. One main difference is that Lowess uses a linear polynomial, while loess uses a quadratic polynomial to fit local subsets of the data.
