Institut National de la Recherche Agronomique (INRA), UR346 Epidémiologie Animale, 63122 Saint Genès Champanelle, France.
Ticks Tick Borne Dis. 2011 Jun;2(2):104-10. doi: 10.1016/j.ttbdis.2011.02.001. Epub 2011 May 8.
In ecology and epidemiology, exploratory field studies based on multivariate statistical models commonly are used to identify factors that are associated with a phenomenon. The challenge is to evaluate whether these factors are indeed correlated to the phenomenon or if the statistical significance results from fortuitous association or type 1 statistical error (probability of rejecting a null hypothesis when it is true). This is particularly the case when the phenomenon is linked to environmental factors that usually are more or less correlated to each other and when the phenomenon is itself highly variable. The abundance of Ixodes ricinus ticks, a major disease vector in Europe, is highly variable and depends on environmental conditions that define suitable habitat and host presence. Our objective was to identify reproducible factors in space and time that influenced the abundance of questing I. ricinus nymphs. We sampled questing nymphs in pastures in 4 settings, i.e. during 3 periods (2003, 2004, and 2006) in one region, and during one period (2006) in another region, both regions located in Central France. The same data collection, data selection, and model analysis using negative binomial distribution were applied independently in the 4 data sets to identify 'reproducible' factors, i.e. explanatory factors that were significant in different time periods and spaces. The 3 most reproducible factors suggested that woodland type vegetation and woodland vicinity constantly favoured nymph abundance on pastures. In addition, the presence of fruit trees was significantly associated with nymph abundance in one region. The other factors were not reproducible. The study confirmed the status of key factors for nymph abundance while avoiding having to redefine the statistical model to model the different sampling conditions. It also shows the difficulty to identify factors with general significance acting on a very variable phenomenon, based on a study made one year in one region. Relevant study design should consider appropriate repetitions based on the range of values of the studied factors. Furthermore, discussion of the results is essential to identify factors that have a general significance.
在生态学和流行病学中，通常基于多元统计模型进行探索性实地研究，以确定与现象相关的因素。挑战在于评估这些因素是否确实与现象相关，或者统计显著性是否来自偶然关联或第一类统计错误（当假设为真时拒绝零假设的概率）。当现象与环境因素相关，而这些环境因素通常或多或少相互关联，并且现象本身高度可变时，尤其如此。欧洲主要疾病媒介壁虱的数量变化很大，取决于确定适宜栖息地和宿主存在的环境条件。我们的目标是确定在空间和时间上影响游离若虫数量的可重复因素。我们在 4 个环境中对牧场中的游离若虫进行了采样，即在一个地区的 3 个时期（2003 年、2004 年和 2006 年）和另一个地区的 1 个时期（2006 年）进行采样，这两个地区都位于法国中部。在 4 个数据集上独立应用相同的数据收集、数据选择和使用负二项分布的模型分析，以确定“可重复”的因素，即不同时间段和空间中显著的解释因素。3 个最可重复的因素表明，林地类型植被和林地附近始终有利于牧场中的若虫数量。此外，在一个地区，果树的存在与若虫数量显著相关。其他因素则不可重复。该研究证实了影响若虫数量的关键因素的地位，同时避免了为模拟不同的采样条件而重新定义统计模型。它还表明，基于在一个地区进行的一年研究，基于研究因素的变化范围，很难确定对高度可变现象具有普遍意义的因素。适当的重复应考虑到相关研究设计，基于所研究因素的变化范围。此外，讨论结果对于确定具有普遍意义的因素至关重要。
