Presentation of smoothers: the family approach (Q1861599)

The authors consider a kernel density estimator for an i.i.d. sample \(X_1\),\dots,\(X_n\), \(f_n(x)=(nh)^{-1}\sum_{i=1}^nK((x-X_i)/h)\), and propose to plot a family of graphs \(f_n(x,h_j)\), \(j=1,\dots,M\). It is really surprising how informative a plot based on this idea can be. The choice of \(M\) and \(h_j\) and the scaling problem are discussed.