{"id":"arxiv/2601.03920","url":"https://arxiv.org/abs/2601.03920","tags":["math.ST"],"score":0,"title":"Adaptive thresholding for wavelet-based nonparametric heteroskedastic variance estimation on the sphere","domain":"math.ST","source":"arxiv","authors":["Claudio Durastanti","Radomyra Shevchenko"],"reviews":0,"subject":"math.ST","summary":"This paper addresses the nonparametric estimation of a spatially varying, heteroskedastic variance function on the unit sphere within a regression framework. While adaptive regression estimation is well-established on manifolds, characterizing localized noise structures presents unique theoretical obstacles due to bias propagation from the unknown mean function. To circumvent this, we propose a fully data-driven, multiresolution estimator based on localized spherical frames, namely, needlets, combined with a hard-thresholding protocol and a sample-splitting scheme. The approach exploits the…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-01-07","reactions":{}}