Statistical Models
Abstract
Efficient planning requires efficient methods for forecasting and control. One of the objects of the present research is to further extend methods described in a recent successful 550 page book by Box and Jenkins developed under AFOSR sponsorship. Non-stationary models which can adequately represent multiple dependent records developing in time have been obtained and efficient methods for identification, estimation and diagnostic checking have been studied. Of particular importance are canonical forms of the model whereby the information cantained in many records can often be summarized in a few composite series. Difficult problems in estimation are being approached using Bayesian methods. Two problems occurring in continuous time control theory were studied and results obtained on the accuracy properties of numerical algorithms for solving them approximately. Some new results in Bayesian Tolerance Regions have been obtained.
Document Details
- Document Type
- Technical Report
- Publication Date
- May 01, 1971
- Accession Number
- AD0735197
Entities
People
- George E. Box
Organizations
- University of Wisconsin Madison Department of Statistics