For nearly eight decades the Luria–Delbrück protocol remains the preferred method for experimentally determining microbial mutation rates. However, earnest development and rigorous applications of statistical methods for mutation rate comparison using fluctuation assay data are a relatively recent phenomenon. While likelihood ratio tests tailored for the fluctuation protocol give investigators appropriate tools, researchers sometimes may prefer to view the comparison of two mutation rates through the lens of the ratio of the two mutation rates. The idea of using the bootstrap technique to construct intervals for mutation rate fold change was proposed nearly a decade ago, but it failed to gain traction partly due to a failure to incorporate likelihood-based estimation. In addition to extending the bootstrap method, this paper proposes two new methods of constructing intervals for mutation rate fold change: a profile likelihood method and a Bayesian method utilizing Monte Carlo Markov chain. All three methods are assessed by large-scale simulations.

近八年来，Luria-Delbrück 协议一直是实验确定微生物突变率的首选方法。然而，使用波动测定数据进行突变率比较的统计方法的认真开发和严格应用是一个相对较新的现象。虽然为波动协议量身定制的似然比测试为研究人员提供了适当的工具，但研究人员有时可能更喜欢通过两种突变率的比率来查看两种突变率的比较。使用 bootstrap 技术构建突变率倍数变化区间的想法是在近十年前提出的，但它未能获得关注的部分原因是未能纳入基于似然的估计。除了扩展bootstrap方法，本文提出了两种构建突变率倍数变化区间的新方法：轮廓似然法和利用蒙特卡罗马尔可夫链的贝叶斯方法。所有三种方法都通过大规模模拟进行评估。