Statistical Assessment of Therapeutic Equivalence Based on Paired Binary Data

• Main Authors: Hseih, Tzung-Cheng, 謝宗成
• Other Authors: Jen-Pei Liu
• Format: Others
• Language: zh-TW
• Published: 1997
• Online Access: http://ndltd.ncl.edu.tw/handle/05292647589635046705

碩士 === 國立成功大學 === 統計學系 === 85 === We consider statistical inference for assessment of therapeutic equivalencebetween two diagnostic procedures based on the paired binary clinical endpointsfor the comparison of diagnostic efficacy with respect to the gold standard indiagnosis of a certain disease. The hypothesis of therapeutic equivalence isformulated as the interval hypothesis based on the difference between correlatedproportions of the correct diagnosis. We propose an asymptotic two one- sidedtests procedure for the interval hypothesis which is also shown to be operation-ally equivalent to the asymptotic confidence interval approach. In addition, anapproximate formula for sample size determination is also suggested. Also, wepropose an exact tests procedure by the method for small sample, and provide alist of critical values for a symmetric equivalence limit of 0.2. A simulationstudy was conducted to empirically examine the size and power of proposedasymptotic tests procedures. Also, we conducted additional simulation study forverifying whether the critical values for exact tests can adequately control thesize. A numerical example will be provided for illustration of the proposedprocedures. We make some final remarks about the critical values for the exacttest and other formulation of the equivalence limits for the interval hypothesis. In addition, the procedure for summarization of the equivalence results overstrata will be also suggested.