Change the degree of the local polynomial in Cleveland's method to 0 (locally constant) using the argument degree (see Comment 8). Using the cars data from Problem 1, how does changing the degree affect the estimates compared to the locally linear estimates (degree=1) from Problem 8? Change the degree to 2 (locally quadratic) and compare graphically the estimate obtained to those found using degrees 0 and 1.
Problem 1
The data set cars from Ezekiel (1930) contains stopping distances for various speeds. Smooth the data using Friedman's smoother by choosing your own value of the span. Use dist as the dependent (response) variable and speed as the independent (predictor) variable. Using trial and error, what seems to be a reasonable span? Comment on the graphical comparison between the estimate using your choice of span with the estimate using the span chosen by cross-validation.
Problem 8
Smooth the data set cars from Problem 1 using Cleveland's smoother. Choose a reasonable value for the span using trial and error. Why does this span appear to be a good choice?
Comment on the graphical comparison of the estimate using the trial and error span with the estimates found using the span chosen by the two cross-validation validation methods (aicc and gcv).