The Annals of Statistics
- Ann. Statist.
- Volume 23, Number 6 (1995), 2218-2240.
Sequential confidence bands for densities
This paper proposes a fully sequential procedure for constructing a fixed-width confidence band for an unknown density on a finite interval and shows the procedure has the desired coverage probability asymptotically as the width of the band approaches zero. The procedure is based on a result of Bickel and Rosenblatt. Its implementation in the sequential setting cannot be obtained using Anscombe's theorem, because the normalized maximal deviations between the kernel estimate and the true density are not uniformly continuous in probability (u.c.i.p.). Instead, we obtain a slightly weaker version of the u.c.i.p. property and a correspondingly stronger convergence property of the stopping rule. These together yield the desired results.
Ann. Statist., Volume 23, Number 6 (1995), 2218-2240.
First available in Project Euclid: 15 October 2002
Permanent link to this document
Digital Object Identifier
Mathematical Reviews number (MathSciNet)
Zentralblatt MATH identifier
Xu, Yi; Martinsek, Adam T. Sequential confidence bands for densities. Ann. Statist. 23 (1995), no. 6, 2218--2240. doi:10.1214/aos/1034713654. https://projecteuclid.org/euclid.aos/1034713654