單一重複添加次序試驗之分散效應篩選

Abstract

添加次序試驗之設計與分析近年來受到許多研究者的重視。目前多數試驗資料分析方法均專注於位置效應之篩選。然而，分散效應亦有可能在添加次序試驗中扮演關鍵的角色。基於置信廣義樞紐量，本論文提出一個適用於單一重複添加次序試驗的分散效應篩選方法。數值模擬結果顯示，新方法能有效地控制型一錯誤發生率。本論文透過分析一筆實際藥物合成試驗說明新方法的可行性，並提供一個交換演算法可用於產生合適的試驗設計。

Recently, statistical designs and analysis methods for order-of-addition experiments have received increasing attention from researchers and practitioners. Most existing methods focus on screening active location effects. However, active dispersion effects may also play a key role in addressing order-of-addition problems. In this study, a new testing procedure is developed using fiducial generalized pivotal quantities to screen out active dispersion effects from unreplicated order-of-addition experiments. Based on the simulation results, the proposed method can control the empirical type I error rates. A drug combination experiment is used to illustrate the proposed method. In addition, a group-wise exchange algorithm is designed to generate eligible designs for testing multiple dispersion effects.