Cross-validated wavelet shrinkage (Q964663)

Different cross-validation (CV) algorithms are considered for threshold selection in wavelet shrinkage regression. In these algorithms the leaving-out sample elements are replaced by some imputed values. Leave-m-blocks out CV and CV for level dependent thresholds are discussed. The CV approaches are compared by an empirical Bayes procedure and complex multiwavelet style hard thresholding via simulations. Analysis of inductance plethysmography data and results of image processing are presented.