Dataset and supplementary information for the article entitled: "Let's DAG in - How DAGs can help Behavioural Ecology be more transparent"

The dataset contains list of articles, whether they contain a DAG or not and the number of variables they used in a single analysis. To quantify the current use of DAGs in behavioural ecology, we reviewed 245 recent articles from two well-known journals in the field- Behavioral Ecology (Volume 34, Issues 4-6 and Volume 35, Issues 1-3 from July/ August 2023 to May/June 2024) and Animal Behaviour (Issues 210-215, April 2024 to September 2024) for their use of DAGs in their methods and as a justification for their statistical analysis. Out of 245 articles, 7 were papers with theoretical modeling studies and hence were not considered, and one paper did not conduct a statistical analysis. We went through the methods, statistical analyses, results and sometimes even the R code to count the maximum number of variables that were used within a single analysis (‘Max_Nr_Variables_in_Single_Analysis’). We counted all the variables within a single model, which was usually a linear mixed model, and we picked the number of variables in the largest model. For eg, y~z1+z2+(1|a1)+(1|a2) has 5 variables within a single model, including the response variable. This is because, to formulate such a statistical model from a DAG, it would need to contain at least 5 variables or more. Therefore, it is a conservative estimate of the total number of variables that behavioural ecologists measure and add to their statistical models. From our review of the articles, we found that, on average, studies used 6 variables in a single statistical regression model, with a range of 2 – 23 variables (as illustrated in the histogram below)本数据集收录了多篇学术文章的条目信息，涵盖每篇文章是否使用了有向无环图（Directed Acyclic Graph，DAG），以及其单次分析中所使用的变量数量。为量化行为生态学领域内有向无环图（DAG）的当前使用现状，我们对该领域两大知名期刊刊载的245篇近期文章开展了综述调研：分别是《行为生态学》（Behavioral Ecology，2023年7/8月至2024年5/6月期间的第34卷第4-6期、第35卷第1-3期）与《动物行为》（Animal Behaviour，2024年4月至9月期间的第210-215期），重点审查其研究方法与统计分析合理性说明中是否使用了DAG。在这245篇文章中，7篇为理论建模类研究论文，故未纳入统计；另有1篇未开展统计分析，同样予以排除。我们通过审阅文章的研究方法、统计分析内容、结果部分，有时甚至包括其配套的R代码，统计得到单次分析中所使用的变量最大数量（记为`Max_Nr_Variables_in_Single_Analysis`）。我们统计了单个统计模型（通常为线性混合模型）内的全部变量，并选取其中变量数量最多的模型的变量数作为统计值。例如，公式`y~z1+z2+(1|a1)+(1|a2)`对应的单次分析中包含5个变量，其中包含响应变量。这是因为，若从DAG出发构建此类统计模型，其所需变量数量至少为5个或更多。因此，该统计值是对行为生态学家实际测量并纳入统计模型的变量总数的保守估计。基于本次文章综述，我们发现：研究人员在单个统计回归模型中平均使用6个变量，变量数量的分布区间为2至23个（详见下文直方图）。