Modeling Episodic Time Series.
Abstract
A particular form of filtered Poisson process is suggested for modeling time series data tht contain episodes, i.e. large, exponentially decaying excursions away from baseline values of the series. The properties of this process are derived and the principle of conditional least squares estimation is used to obtain consistent, asymptotically normal estimators of the average excursion heights and the decay rate parameter. The method is illustrated using hormone levels data. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 01, 1982
- Accession Number
- ADA129729
Entities
People
- H. Joseph Newton
Organizations
- Texas A&M University