Incident data that occur in close spatial and temporal proximity often share latent or unobserved influences. Understanding the spatiotemporal associations among different categories of such incidents is therefore crucial for urban studies and public health research. Spatial statistical methods have been widely employed to investigate association patterns; however, several methodological challenges persist – particularly regarding sample size determination, class imbalance, and bandwidth selection. This study proposes a methodological framework for analyzing multi-temporal scale association patterns in incident data using the Geographically and Temporally Weighted Co-Location Quotient (GTWCLQ) method. First, we design and validate a systematic parameter optimization approach to address limitations in sample size, class distribution, and spatial-temporal bandwidth settings. Second, we develop a structured framework to explore the spatiotemporal associations across multiple temporal scales in the incident data. We demonstrate the utility of this framework through an empirical case study examining the spatiotemporal association patterns of childhood respiratory diseases in Nanning City, China, using incident data from December 2016 at both monthly and daily resolutions. The results reveal that our validated multi-scale spatiotemporal association analysis framework effectively captures the dynamic associations in disease incident data across different temporal scales, visualizes the spatiotemporal heterogeneity, and further examines the scaling effect of multiple temporal data on the co-location patterns. The findings contribute to methodological rigor in co-location association analysis of spatiotemporal incident data and have practical implications for disease surveillance, environmental health monitoring, and spatial decision-making.
LI et al. (Sun,) studied this question.