Accounting for historical information in designing experiments: the Bayesian approach