Bayesian Prediction With Linear Dynamic Model: Principle And Application

In the business applications where only a few data is observed, statistical models estimated in frequentist framework is not reliable or even not obtainable. Bayesian updating, by calculating subjective probabilities conditional on real observations, could form optimal prediction given some prior belief. Through a demonstration of cash flow prediction example, the Bayesian method and a frequentist method, ordinary least square (OLS) to be specific, are compared. Bayesian model has similar performance as OLS in the example and moreover provides a solution to the situations where OLS is inapplicable.