Highly Informative Priors

SUMMARY After discussing the role of prior information in statistical inference, historically and in current problems, we analyze the problem of seasonal adjustment i n economics. Litterman 1980 has shown how informative priors for autoregressive coeecients can improve economic forecasts. We nd that in seasonal adjustment informative priors can have a m uch greater eeect on our conclusions. In our model, even the dimensionality of the joint posterior distribution of the irregulars depends on prior information about the seasonal component; and some functions of the ir-regulars can be determined more accurately than in sampling theory.