MAXIMUM-LIKELIHOOD-ESTIMATION UNDER ORDER RESTRICTIONS BY THE PRIOR FEEDBACK METHOD

Algorithms for deriving isotonic regression estimators id order-restri cted linear models and more generally restricted maximum likelihood es timators are usually quite dependent on the particular problem conside red. We propose here an optimization method based on a sequence of for mal Bayes estimates whose variances converge to zero. This method, aki n to simulated annealing, can be applied ''universally''; that is, as long as these Bayes estimators can be derived by exact computation or Markov chain Monte Carlo sampling approximation. We then give an illus tration of our method for two real-life examples.