{"id":"arxiv/2609.29231","url":"https://arxiv.org/abs/2609.29231","tags":["math.ST"],"score":0,"title":"Choosing the penalty in nonparametric regression: short and long-range dependence","domain":"math.ST","source":"arxiv","authors":["Emmanuel Caron Parte","Jérôme Dedecker","Bertrand Michel"],"reviews":0,"subject":"math.ST","summary":"In this work, we study the one-dimensional regression problem under random design and Gaussian errors. Our framework is very general: we make no prior assumptions about the design (which may be nonstationary and exhibit short or long-range dependence), nor do we assume that the errors are homoscedastic. We examine in detail the cases where the error process exhibits short or long-range dependence. We adopt a least-squares penalized strategy using piecewise polynomials to estimate the regression function, following the framework of Baron, Birg{é} and Massart [1999]. We derive explicit…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-09-24","reactions":{}}