A Natural Approach for Detecting Causal Relationship in Time Series.
Abstract
A definition of causality given by C.W.J. Granger lends itself to statistical hypothesis testing for identifying the causal relationship between two time series. The appropriate statistical analysis to be used, however, is not apparent from the definition. One's first impression is to regress present values of one time series on past values of both time series. There are, however, several weaknesses inherent in this procedure. Natural modifications of this procedure lead eventually to a Box-Jenkins approach for suitably preprocessing the time series followed by a regression analysis of the residuals, and analysis which identifies the causal structure between the two time series. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 23, 1976
- Accession Number
- ADA048529
Entities
People
- Frederick W. Morgan
- M. Edwin Ireland
Organizations
- Clemson University