Regression splines provide a flexible family of curves that can be fit to data using ordinary least squares linear model methods, yet closely model arbitrary smooth response curves. We study how B-spline and natural cubic spline basis functions allow us to do this. We also discuss how to use k-fold cross-validation to select the optimal number of knots or degrees of freedom for the basis. As an illustration the methods are applied to birth rate data from 1917 to 2003.
…Read more
Less…