* Example of nonparametric regression adapted from * Rust, R.T., "Flexible Regression", Journal of Marketing Research, * Vol. 25, 1988, pp. 10-24. * * The experiment is designed to show the limitations of OLS when the * relationship is nonlinear. * SAMPLE 1 50 SET RANFIX * Generate the data using the same experiment given in Rust (1988) GENR X=UNI(1) GENR E=NOR(.1) GENR TRUE = 1 - 4*(X-.5)**2 GENR Y = TRUE + E * Try OLS OLS Y X / ANOVA PREDICT=YOLS * Nonparametric regression with a normal kernel function NONPAR Y X / PREDICT=YHAT * Now compare the results PRINT TRUE YOLS YHAT * Plot the results SORT X TRUE YOLS YHAT GRAPH TRUE YOLS YHAT X / LINEONLY STOP