Statistics Meeting Book (February 20, 2019)

Goals • Principal question is size of dPOD(M1,M2) and its 95% confidence interval • If zero is contained within the 95% C.I. for dPOD, then have little evidence from study to support claim that methods differ • Want 95% C.I. estimator that is accurate (within 84.5% to 98.5% coverage) across a wide range of conditions • Formula should be ‘Excelable’ copyright 2016 by Least Cost Formulations, Ltd., all rights reserved. 7 Status quo ante • Previous estimator for dPOD was based on paired differences d i for each test portion replicate. • d i = 1 if (M1, M2) = (+, –) • d i = 0 if (M1, M2) = (+, +) or (-, -) • d i = -1 if (M1, M2) = (-, +) • dPOD = average d i • C.I. is student t interval on d i ’s copyright 2016 by Least Cost Formulations, Ltd., all rights reserved. 8