Distribution-free lower bounds in density estimation (Q800659)

We consider the kernel estimate on the real line, \[ f_ n(x)=(nh)^{- 1}\sum^{n}_{i=1}K((X_ i-x)/h), \] where K is a bounded even density with compact support, and \(X_ 1,...,X_ n\) are independent random variables with common density f. We treat the problem of placing a lower bound on the L1 error \(J_ n=E(\int | f_ n-f|)\) which holds for all f. In particular, we show that there exist \(A(K)\geq (9/125)^{1/5}\) depending only upon K, and \(B^*(f)\geq 1\) depending only upon f such that (i) for all \(f: \inf_{h>0}n^{2/5}J_ n\geq CA(K)B^*(f)+o(1)\geq 0.6076703...+o(1)\), where \(C=1.028493..\). is a universal constant; (ii) for all f with compact support and two bounded continuous absolutely integrable derivatives, \(\inf_{h>0}n^{2/5}J_ n\leq C^*A(K)B^*(f)+o(1)\), where \(C^*=1.3768102..\). is another universal constant. For this class of densities, we also obtain the exact asymptotic behavior of \(J_ n\).