Flexible data modelling via skew mixture models:challenges and applications

该项目旨在探索处理非正态特征数据的新模型。参数分布是统计建模和推理的基础。几个世纪以来，“正态”分布一直是连续数据的主导模型。然而，实际数据很少满足正态性假设。因此，对更灵活的发行版有强烈的需求。该项目旨在开发使用偏分量分布的有限混合建模的新方法，为处理具有非正态特征（如偏度、重/轻尾和多模态）的数据提供更好的模型。应用可能包括安全入侵检测，临床诊断和预后，以及流式细胞术和质量细胞术。

This project seeks to explore new models for handling data with non-normal features. Parametric distributions are fundamental to statistical modelling and inference. For centuries, the ‘normal’ distribution has been the dominant model for continuous data. However, real data rarely satisfy the assumption of normality. There is thus a strong demand for more flexible distributions. This project aims to develop new methodologies in finite mixture modelling using skew component distributions to provide better models for handling data with non-normal features (such as skewness, heavy/light tails, and multimodality). Applications may include security intrusion detection, clinical diagnosis and prognosis, and flow and mass cytometry.