Data-based optimal smoothing of a wavelet density estimator (Q1381769)

The paper considers the problem of estimating a probability based on a complete random sample, using a wavelet-based orthogonal expansion. Section 2 introduces the concepts of wavelets and a basic wavelet density estimator. Section 3 defines the modifications on the empirical wavelet coefficients to construct the new proposed wavelet density estimator, and develops a method for estimating the optimal smoothing parameters based on a finite random sample. Section 4 contains a numerical experiment supporting the promising performance of the modified estimator, while its asymptotic analysis shows for it optimal rates of convergence.