SYNOPTIC ABSTRACT
Consider the problem of estimating a linear combination of means from populations with different, unknown variances. We study the Bayes version of the problem and derive the dynamic programming optimality equations for the determination of optimal multi stage sequential sample size allocation procedures for several pertinent loss structures. We point out the relation of these equations to the “inventory equations”. We develop heuristic procedures for the case of Bernoulli and normal populations and give numerical comparisons.
Key Words and Phrases: