{"id":"arxiv/2610.05511","url":"https://arxiv.org/abs/2610.05511","tags":["math.ST","eess.SP","stat.ME"],"score":0,"title":"Rank-one variance inflation in polynomial estimation of signal parameters with a sign-preserving fractional-power basis","domain":"math.ST","source":"arxiv","authors":["Serhii Zabolotnii"],"reviews":0,"subject":"math.ST","summary":"The polynomial maximization method estimates signal parameters in non-Gaussian noise from a finite description of the noise by its moments and cumulants, without a density. Known accuracy results assume that these noise characteristics are known, or use a design in which estimating them jointly costs no accuracy. In practice they are estimated from the residuals of a preliminary fit, and the effect of this substitution on accuracy is unknown. We treat the feasible scheme as a two-step estimator and, for a general smooth signal model in a fixed design, find the increment of its asymptotic…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-10-04","reactions":{}}