In my earlier paper entitled "Bayesian Analysis - Applications to Danish Data" marginal posterior probability distributions for parameters in two regression models were analysed. In this article we shall be concerned with the analysis of conditional posterior probability distributions for parameters under given conditions. Two linear regression models are considered for the purpose. Under the assumption of autocorrelated errors and given initial conditions, the conditional posterior probability distribution for marginal propensity to consume parameter is practically insensitive to negative autocorrelation, but it is very sensitive to a high positive autocorrelation. Thus, least squares estimates of parameters in a regression model with a strong departure from independence of errors may lead us to wrong conclusions. However, the effect of a change in the initial value of the dependent variable on the conditional distribution is meoderate depending upon the extent of variation in the data.
