Abstract DeltaSA is an R-package and a Java on-line tool developed at the EC-Joint Research Centre to assist and benchmark source apportionment applications. Its key functionalities support two critical tasks in this kind of studies: the assignment of a factor to a source in factor analytical models (source identification) and the model performance evaluation. The source identification is based on the similarity between a given factor and source chemical profiles from public databases. The model performance evaluation is based on statistical indicators used to compare model output with reference values generated in intercomparison exercises. The references values are calculated as the ensemble average of the results reported by participants that have passed a set of testing criteria based on chemical profiles and time series similarity. In this study, a sensitivity analysis of the model performance criteria is accomplished using the results of a synthetic dataset where “a priori” references are available. The consensus modulated standard deviation punc gives the best choice for the model performance evaluation when a conservative approach is adopted.

中文翻译：

---

用于源分配基准、描述和敏感性分析的 DeltaSA 工具

摘要 DeltaSA 是由 EC 联合研究中心开发的 R 包和 Java 在线工具，用于辅助和基准源分配应用程序。其关键功能支持此类研究中的两项关键任务：在因子分析模型中将因子分配给源（源识别）和模型性能评估。来源识别基于给定因素与来自公共数据库的来源化学概况之间的相似性。模型性能评估基于用于将模型输出与比对练习中生成的参考值进行比较的统计指标。参考值计算为参与者报告的结果的整体平均值，这些结果已通过一组基于化学特征和时间序列相似性的测试标准。在这项研究中，模型性能标准的敏感性分析是使用合成数据集的结果完成的，其中“先验”参考可用。当采用保守方法时，共识调制标准偏差 punc 为模型性能评估提供了最佳选择。