Comprehensive comparisons of major sequential design procedures for sensitivity testing

The main goal of a good sensitivity testing procedure is to quickly determine at which stimulus level a specified fraction of test items fail. Based on a previous study, two procedures, D-optimal and 3pod, appear to be most promising. In this paper we expand on this work, first by considering other variants of the two procedures. A related procedure to D-optimal, called Sen-Test, is examined and shown to be superior to D-optimal in a simulation study. A revised version of 3pod, dubbed 3pod2.0, is proposed. A comprehensive simulation study to compare Sen-Test, 3pod and 3pod2.0 is conducted. Based on this study, we draw some concluding remarks on their relative performance and make recommendations.

优质灵敏度测试流程的核心目标，在于快速确定当刺激强度达到何种水平时，指定比例的受试样本会发生失效。基于既往研究，D最优（D-optimal）与3pod两种测试流程被视为最具应用潜力的方案。本文在此前研究的基础上进行拓展：首先针对这两种流程的各类变体展开探讨；其次对与D最优相关的Sen-Test测试流程进行考察，经仿真研究证实，该流程的性能优于D最优。同时，本研究提出了3pod的修订版本，命名为3pod2.0。随后开展了一项全面的仿真实验，对Sen-Test、3pod与3pod2.0三种流程的性能进行对比。基于本次实验结果，本文就三种流程的相对性能给出总结性评述，并提出相关建议。