Hot spot identification method based on Andrews curves: an application on the COVID-19 crisis effects on caregiver distress in neurocognitive disorder

Identifying and locating areas – hot spots – that present high concentration of observations in a high-dimensional data set is crucial in many data processing and analysis methods and techniques, since observations that belong to the same hot spot share information and behave in a similar way. A useful tool towards that aim is the reduction of the data dimensionality and the graphical representation of them. In the present paper, a new method to identify and locate hot spots is proposed, based on the Andrews curves. Simulations results demonstrate the performance of the proposed method, which is also applied to a high-dimensional data set, regarding caregiver distress related to symptoms of people with neurocognitive disorder and to the mental effects of the recent outbreak of the COVID-19 pandemic.

在诸多数据处理与分析方法及技术中，识别并定位高维数据集中观测值高度聚集的区域——热点区域（hot spots）是一项至关重要的任务，这是因为同属一个热点区域的观测值会共享信息且行为模式高度相似。实现该目标的一项实用工具是数据降维与数据可视化。本文提出了一种基于安德鲁斯曲线（Andrews curves）的热点区域识别与定位新方法。仿真结果验证了所提方法的性能，同时将该方法应用于一项高维数据集，该数据集涵盖两大主题：一是与神经认知障碍（neurocognitive disorder）患者症状相关的照料者心理困扰，二是近期新型冠状病毒肺炎（COVID-19）大流行带来的精神健康影响。