Some Approaches to Testing Hypotheses for Multidimensional Stationary Stochastic Processes.
Abstract
Several techniques for testing hypotheses concerning multidimensional stationary stochastic processes have been developed. These were applied to the two-dimensional discrete finite Fourier transforms of space-time series. The justification for this is the asymptotic normality and independence of the transformed variables. When the space-time series under consideration consists of two-dimensional signal functions imbedded in and common to all realizations of a stationary noise process, a likelihood ratio test can be applied in the transformed domain. If the signal model includes an unknown epoch or phase which varies from realization to realization, no true replications are possible, and the test must be modified. If one is willing to pay the price of increased computation and increased errors of both kinds, the modified test has reasonable power at acceptably low test levels. However, an ad hoc test is at least as powerful at all test levels and is considerably more powerful at very low levels.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 02, 1972
- Accession Number
- AD0749687
Entities
People
- David A. Swick
Organizations
- United States Naval Research Laboratory