Classification of Time-Dependent Observations
Author: Azen, S. P.
Corporate Author: RAND Corporation
Corporate Report Number: R-471-PR
Date of Publication: 1969-08
Pages: 112
Contract: F44620-67-C-0045
Project: Project RAND
AD Number: AD0692658
Abstract: The report provides a discussion of the problem of classifying a series of time-dependent correlated observations into one of several populations. Various procedures are investigated under the assumptions that the time series are approximated by polynomials in t, and that the observations have a special covariance structure. Relative efficiencies of the various procedures are compared. The model is shown to be realistic for applications in certain psychological and physiological measurement problems in which an investigator takes measurements on an individual for classification purposes.
Provenance: Borg-Warner