Keele Research Repository

The thesis deals with binomial and multinomial sequential selection problems. Optimal sequential sampling schemes are derived by using dynamic programming in conjunction with loss functions and sampling costs and to minimize expected sample sizes. Comparisons are carried out with sampling schemes where observations are taken in blocks or groups and with fixed sample size procedures.
Several suboptimal designs are suggested and numerical comparisons are made under several performance characteristics which are obtained exactly and by using Monte Carlo simulation. The performance of the procedures is studied when the parameters are fixed and where they are generated from particular prior distributions.