Sankhya: The Indian Journal of Statistics

2006, Volume 68, Pt. 4, 569--586

Exact Wavelet Designs for Discrimination

Yi Liu, Memorial University of Newfoundland, Canada
Alwell J. Oyet, Memorial University of Newfoundland, Canada

SUMMARY. We discuss the construction of minimax Bayesian T-optimal, D-ratio optimal and sequential G-optimal designs for discrimination between competing wavelet representations of nonparametric regression models. In our examples we use the multiwavelet and Daubechies wavelet systems. We find that symmetry is an important property of designs for multiwavelet models, whereas designs for Daubechies wavelet models are nonsymmetric.

AMS (2000) subject classification. Primary 62G08; secondary 62K05.

Key words and phrases. Daubechies wavelet, D-optimality, multiwavelet, nonparametric regression, T-optimality, sequential designs, weighted least squares.

Full paper (PDF)