Nonparametric Estimation of Signals Mixed with Noise.
Abstract
Analysis of a set of evolutionary or nonstationary time series data is traditionally carried out by the use of regression and spectral methods. Although these procedures are, to some extent, nonparametric in nature, the basic assumption implicit in such data analysis has usually been that the time series is Gaussian or nearly so. We do not know very well how efficient and useful these procedures are in case the data structure is distinctly non-Gaussian. Classical data analysts who handle data sets composed of independent observations have noticed over the last three decades that robust and adaptive nonparametric methods have worked very efficiently even for situations where large implicit variability in the data has effectively ruled out a Gaussian model. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Feb 01, 1982
- Accession Number
- ADA113018
Entities
People
- Kamal C. Chanda
Organizations
- Texas Tech University