Sampling Designs for Estimating Integrals of Stochastic rocesses Using Quadratic Mean Derivatives
Abstract
The problem of estimating the integral of a stochastic process by linear estimators based on observations of the process and its existing quadratic mean (q.m.) derivatives at a finite number of sampling points is considered. The process is assumed to have exactly K q.m. derivatives, K=O, 1,2, ..... The asymptotic performance of optimal-coefficient estimators that depend on an inverse matrix is determined for regular sampling designs under slightly different assumptions than those in Sacks and Ylvisaker (1970). Simple- coefficient estimators based on a trapezoidal rule with a correction term that depends on the q.m. derivatives of the process at all sampling points of a regular design are introduced. Their asymptotic performance is identical to that of the optimal-coefficient estimators. Keywords: Linear equations, Stochastic processes, Integrals.
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 01, 1990
- Accession Number
- ADA225961
Entities
People
- Karim Benhenni
- Stamatis Cambanis
Organizations
- University of North Carolina at Chapel Hill