Abstract:
Fish photolocomotor behavioral response (PBR) studies have become increasingly prevalent in pharmacological and toxicological research to assess the environmental impact of various chemicals. There is a need for a standard, reliable statistical method to analyze PBR data. The most common method currently used, univariate analysis of variance (ANOVA), does not account for temporal dependence in observations and leads to incomplete or unreliable conclusions. Repeated measures ANOVA, another commonly used method, has drawbacks in its interpretability for PBR study data. Because each observation is collected continuously over time, we instead consider each observation to be a function and apply functional ANOVA (FANOVA) to PBR data. Using the functional approach not only accounts for temporal dependency but also retains the full structure of the data and allows for straightforward interpretation in any subregion of the domain. Unlike the traditional univariate and repeated measures ANOVA, the FANOVA that we propose is nonparametric, requiring minimal assumptions. We demonstrate the disadvantages of univariate and repeated measures ANOVA using simulated data and show how they are overcome by applying FANOVA. We then apply one-way FANOVA to zebrafish data from a PBR study and discuss how those results can be reproduced for future PBR studies.
摘要:
鱼类的光运动行为反应(PBR)研究在药理和毒理学研究中越来越普遍,以评估各种化学物质对环境的影响。需要一个标准,可靠的统计方法来分析PBR数据。目前最常用的方法,单变量方差分析(ANOVA),不考虑观测中的时间依赖性,并导致不完整或不可靠的结论。重复测量方差分析,另一种常用的方法,在PBR研究数据的可解释性方面存在缺陷。因为每个观察都是随着时间的推移连续收集的,相反,我们将每个观察值视为一个函数,并将功能方差分析(FANOVA)应用于PBR数据。使用函数方法不仅考虑了时间依赖性,而且还保留了数据的完整结构,并允许在域的任何子区域中进行直接解释。与传统的单变量和重复测量方差分析不同,我们提出的FANOVA是非参数的,需要最少的假设。我们使用模拟数据证明了单变量和重复测量方差分析的缺点,并展示了如何通过应用FANOVA来克服这些缺点。然后,我们将单向FANOVA应用于PBR研究中的斑马鱼数据,并讨论如何为未来的PBR研究重现这些结果。