Robust designs for Haar wavelet approximation models

摘要

In this paper, we discuss the construction of robust designs for heteroscedastic wavelet regression models when the assumed models are possibly contaminated over two different neighbourhoods: G_1 and G_2. Our main findings are: (1) A recursive formula for constructing D-optimal designs under G_1; (2) Equivalency of Q-optimal and A-optimal designs under both G_1 and G_2; (3) D-optimal robust designs under G_2; and (4) Analytic forms for A- and Q-optimal robust design densities under G_2. Several examples are given for the comparison, and the results demonstrate that our designs are efficient.