A FLEXIBLE AND PARSIMONIOUS MODELLING STRATEGY FOR CLUSTERED DATA ANALYSIS

• 成果类型: Article
• 署名作者: Huang, Tao; Pei, Youquan; You, Jinhong; Zhang, Wenyang
• 署名单位: Shanghai University of Finance & Economics; Shandong University; University of Macau
• 刊物名称: ANNALS OF APPLIED STATISTICS
• ISSN/ISSBN: 1932-6157
• DOI: 10.1214/25-AOAS2018
• 发表日期: 2025
• 页码: 1362-1381
• 关键词: feature-selection; moving average; coefficient; identification

摘要

The development of an appropriate statistical modelling strategy is of paramount importance for the successful analysis of data. The trade-off between flexibility and parsimony is of vital importance in statistical modelling. In the context of clustered data analysis, it is essential to account for the inherent heterogeneity between clusters while simultaneously ensuring parsimony to mitigate the potential for complexity and to preserve the homogeneity within clusters. The objective of this paper is to propose a flexible and parsimonious modelling strategy for clustered data analysis. The strategy strikes an optimal balance between flexibility and parsimony, effectively accounting for both heterogeneity and homogeneity among the clusters, which often possess significant practical implications. In particular, we apply this modelling strategy to analyse data related to the spread of COVID-19 in China, examining how factors such as human mobility and temperature can capture dynamic and nonlinear patterns in the data. We develop an estimation procedure for the unknown parameters, establish the asymptotic properties of the estimators, and conduct simulation studies to evaluate the performance of our method. Additionally, the real data analysis illustrates the practical application of our approach in understanding regional differences in epidemic spread, with implications for public health policy. This flexible framework also offers insights for transfer learning scenarios beyond clustered data analysis.