Default Bayesian Priors for Regression Models with First-Order Autoregressive Residuals

The objective of this paper is to develop default priors when the parameter of interest is the autocorrelation coefficient in normal regression models with first-order autoregressive residuals. Jeffreys' prior as well as reference priors are found. These priors are compared in the light of how accurately the coverage probabilities of Bayesian credible intervals match the corresponding frequentist coverage probabilities. It is found that the reference priors have a definite edge over Jeffreys' prior in this respect. Also, the credible intervals based on these reference priors seem superior to similar intervals based on certain divergence measures. Copyright 2003 Blackwell Publishing Ltd.