BAYESIAN ANALYSIS OF THE REGRESSION MODEL WITH AUTOCORRELATED ERRORS.
Technical rept. no. 22,
WISCONSIN UNIV MADISON
Pagination or Media Count:
Bayesian methods are used to analyze the regression model with errors generated by a first order auto-regressive scheme. For a simple regression model, a derivation is made of finite sample joint, conditional and marginal posterior distributions of the parameters of the model. With these distributions, inferences can be made about parameters and investigations can be made as to how departures from independence, very often encountered in economic data, affect inferences about parameters. Further, this approach provides a unified treatment of non-explosive and explosive models and in fact yields results for deciding whether a process is or is not explosive. To illustrate application of the techniques, two sets of artificially generated data, one set from a nonexplosive model and the other from an explosive model, are analyzed in detail. Then, techniques are developed for a Bayesian analysis of the multiple regression model with autocorrelated errors. Author