Show that if one replaces
1 0 |f(k) (x)| 2 dx
by
R L |f(k) (x)| 2 dx
in the definition of the penalized least squares estimate for some
−∞ ≤ L ≤ min{X1,...,Xn} ≤ max{X1,...,Xn} ≤ R ≤ ∞,
then the values of the estimate on [min{X1,...,Xn}, max{X1,...,Xn}] do not change. Hint: Apply Lemma 20.2 with
a = min{X1,...,Xn} − and b = max{X1,...,Xn} +
and show that a function f ∈ Ck(R), which minimizes
1 n n i=1 |f(Xi) − Yi| 2 + λ
R L |f(k) (x)| 2 dx,
satisfies f(k) (x) = 0 for L<>