SEARCH

SEARCH BY CITATION

Keywords:

  • Bayesian regression;
  • informative priors;
  • predictive densities

This paper considers the problem of undertaking a predictive analysis from a regression model when proper conjugate priors are used. It shows how the prior information can be incorporated as a virtual experiment by augmenting the data, and it derives expressions for both the prior and the posterior predictive densities. The results obtained are of considerable practical importance to practitioners of Bayesian regression methods.