Sieves for Gaussian Processes.
Abstract
The PI has established several important properties of a proposed sieve estimator for the mean of a Gaussian process of known covariance: The estimator is itself a Gaussian process; under a separability assumption, it is asymptotically unbiased and weakly consistent at each time t, and is strongly consistent (globally) in an appropriate norm. For a Gaussian process with zero mean and unknown covariance, the PI has shown that the likelihood for the covariance is in general unbounded almost surely. Moreover, he has developed properties of a proposed sieve estimator for the covariance analogous to those for the mean. No assumption is made about the nature of the time parameter t, either for the mean estimator or for the covariance estimator. Keywords: Hilbert space; Maximum likelihood estimation. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Nov 27, 1985
- Accession Number
- ADA166055
Entities
People
- Jay H. Beder
Organizations
- University of Wisconsin–Milwaukee