Post-Data Two Sample Tests of Location.

An approach to two-sample tests of location may be based on the conditional distribution of the sampling process given that the observations are known constants. The author of this article considers a two-sample problem in which a random sample is drawn and an independent random sample is the cumulative probability functions of two continuous random variables. Tests are constructed for a difference in central tendency of the two distributions where central tendency is indicated in various ways. The test procedures are derived as post-data tests and are based on the Dempster-Shafer theory. The description and methods follow economist Arthur P. Dempster more closely since his work is more related to the problems under study. According to the author, the motivation for this article has been to show that Dempster's theory of inference may be easily applied to nonparametric problems. The presentation is kept to its simplest form, and he does not introduce the multivalued mapping necessary to handle discrete problems.