Statistical Methods In Toxicology

在开发用于分析来自毒理学研究的多变量数据的非参数贝叶斯方法方面取得了进展。这些方法使人们能够评估随着暴露于潜在有害化学物质的水平的变化，健康反应的概况随着时间的变化而发生的变化。为了让不同的动物有不同的不良反应倾向，例如肿瘤的发展，一个常见的建模策略是包括一个潜在的疲劳变量。我们的贡献是允许这种脆弱性随着动物年龄的增长而灵活变化。这一脆弱的轨迹可以被认为是随着年龄的增长而发生的基因表达变化和关键基因突变的低维总结。我们将这一框架应用于癌症化学预防实验中的数据，这些实验调查了水果和蔬菜中抗氧化剂的效果。

Work has progressed in developing nonparametric Bayes methods for analysis of multivariate data from toxicology studies. These methods allow one to assess how the profile of health responses over time changes as the level of exposure to a potentially adverse chemical changes. In order to allow individual animals vary in their tendency to have an adverse response, such as the development of a tumor, a common modeling strategy is to include a latent fraily variable. Our contribution is to allow this frailty to vary flexibly over time as the animal ages. This frailty trajectory can be considered as a low dimensional summary of the changes that occur with age in gene expression and the occurrence of mutations in key genes. We applied this framework to data from cancer chemopreventive experiments investigating the effects of antioxidants in fruit and vegetables.