在無參數混合效應模型中進行由資料驅使的碎形波收縮

Abstract

這篇論文利用觀察到的資料來選擇在無參數混合效應模型中的碎型波收縮的參數,方法包括交叉確認法(cross-validation method)與廣義的交叉確認法(generalized cross-validation)兩類方法.經由模擬研究,我們發現,特別是在樣本數較小及誤差程度不知道時,廣義的交叉確認法是較有效的,較可靠的,及較快的方法.

This study investigates the data-driven selection method of the threshold parameter in the wavelet shrinkage method for nonparametric mixed-effects models. The cross-validation and generalized cross-validation methods are studied. Through the simulation studies, the generalized cross-validation method turns out to be a efficient, reliable, and fast method in varticular when sample sizes are small and the noise levels are unknown.