Panels and Time Series Analysis: Markov Chains and Autoregressive Processes
Abstract
Statistical inference in two time series models is developed for cases where there are several observations on the entire time series. Emphasis is on first-order processes: the Markov chain for discrete data and the first- order autoregressive process for vectors of continuous variables. The models are not necessarily homogeneous (or stationary) in time. Sufficient statistics and maximum likelihood estimates are presented. (Continuous variables are assumed normally distributed.) Test criteria for various hypotheses are developed; on a large-sample basis these criteria have chi square-distributions. The close correspondence between properties and statistical methods for the two models is pointed out.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 01, 1976
- Accession Number
- ADA030653
Entities
People
- Theodore W. Anderson
Organizations
- Stanford University