Key points are not available for this paper at this time.
Summary When a choice has to be made in favour of one of two populations the cost of sampling (experimenting) in order to obtain information on which to base the decision must be balanced against the cost of making the wrong choice. Wald’s (1950) minimax procedure is used to determine a sequential method of sampling when a decision is to be made between two normal populations on the basis of their sample means. It is assumed that the cost of making a wrong choice is given by the difference of the population means multiplied by a factor representing the use made of the chosen population. The population variances are taken as known and equal. It is found that this determines a sequential procedure defined by the ratio of the difference it is desired to detect to a function of the probability with which it is desired to detect that difference, therefore corresponding to a range of tests in which these constants are fixed for the statistician by the user of the results. The performance of this procedure is compared with that for a fixed sample size and found to result in smaller losses over all values of the difference between the populations. A comparison is also made with the double sample procedure of Grundy et al. (1956) on which it is also generally an improvement.
Rita J. Maurice (Mon,) studied this question.