MULTISCALE STATISTICAL ANALYSIS FOR INVERSE PROBLEMS IN CORRELATED NOISE
Marc Raimondo, Justin Wishart · 2009
Existing methods of nonparametric estimation include (among others) classical kernel methods, orthogonalseries approach and wavelet thresholding algorithms. In particular, in case of long memory errors, the kernelmethod was studied in [4] (fixed-design regression) and [5] (random-design regression). The wavelet thresh-olding was studied in [14] (fixed-design case). One has to mention at this point, that in case of long memoryerrors, fixed-design and random-design regression has to be treated in a completely different way, unlike incase independent, identically distributed (i.i.d.) errors.The fixed-design nonparametric regression is often referred as