% Local Regression and Likelihood, Figure 8.3. % % Discrimination/Classification, iris data for different % smooting paramters. % % Note that the iris.mat file contains the full iris dataset; % only Versicolor and Virginica are used in this example. % % The `Species' variable contains species names. `Specn' has % them numerically, Setosa=1, Versicolor=2, Virginica=3. % % Author: Catherine Loader % % Need: get contour plot correct. % Need: distinguish colors in plot. load iris; a = (2:9)/10; z = zeros(size(a)); u = find(Specn >= 2); pw = PetalWid(u); pl = PetalLen(u); y = (Specn(u)==3); for i = 1:length(a) fit = locfit([pw pl],y,'deg',1,'alpha',a(i),'ev','cros','scale',0,'family','binomial'); fv = fitted(fit); tb = tabulate(10*y+(fv>=0.5)) z(i) = sum(y == (fv<=0.5)); end; [a; z] fit = locfit([pw pl],y,'deg',1,'scale',0,'family','binomial'); figure('Name','fig8_3: iris classification'); lfplot(fit,'contour');