An exact confidence interval from untransformed data for the ratio of two formulation means

Abstract

Suppose that a test formulation and a standard formulation are to be compared in an experiment with a two period crossover design. Suppose that it is desired to obtain a confidence interval for the ratio of the test formulation mean to the standard formulation mean for some variable. Suppose that the measurements on the variable will themselves be analyzed rather than analyzing the logarithms of the measurements. A method for obtaining an exact confidence interval for a rather general model is described. A method has previously been given for obtaining an exact confidence interval under a more restrictive model. The two models are contrasted.