Exploring Multiscale Spatial Interactions: Multiscale Geographically Weighted Negative Binomial Regression

Abstract

In this article, I develop and implement the multiscale geographically weighted negative binomial (MGWNB) model, extending the spatially weighted interaction models by integrating a multiscale framework. This model effectively tackles the multiscale nonstationarity and overdispersion issues found in spatial interaction models. By comparing it with multiscale geographically weighted Poisson regression using simulated data, I demonstrate its superior performance in several aspects, including its capability to estimate the scale of processes, its effectiveness in capturing the spatial heterogeneity, and its ability to produce a better goodness of fit. The application of MGWNB in interprovincial population migration in China, using 2020 Chinese census data, also demonstrates its effectiveness and efficiency, revealing strong multiscale spatial heterogeneity in the migration patterns.

本文研究并实现了多尺度地理加权负二项模型(MGWNB), 通过采用多尺度框架, 拓展了空间加权交互模型。该模型有效地解决了空间交互模型的多尺度非平稳性和过分散问题。根据模拟数据, 对比了MGWNB模型和多尺度地理加权泊松回归, 证明了MGWNB的几个优势: 能估计分析的尺度、能有效地捕捉空间异质性、有更好的拟合度。利用2020年中国人口普查数据, MGWNB在中国跨省人口流动中的应用也证明了其有效性和效率, 揭示了流动模式中的多尺度强空间异质性。

En este artículo, desarrollo y aplico el modelo binomial negativo y multiescalar geográficamente ponderado (MGWNB), extendiendo los modelos de interacción espacialmente ponderados, para integrar un marco multiescalar. Este modelo aborda efectivamente las cuestiones de la no estacionalidad multiescalar y de sobredispersión que se hallan en los modelos de interacción espacial. Con su comparación con la regresión Poisson geográficamente ponderada a multiescala, usando datos simulados, demuestro su mejor rendimiento en varios aspectos, incluyendo su capacidad para calcular la escala de los procesos, su eficacia en captar la heterogeneidad espacial y su habilidad para producir un mejor resultado en el ajuste. La aplicación de MGWNG en la migración interprovincial de la población en China, usando datos del censo chino de 2020, demuestra también su efectividad y eficiencia, revelando una fuerte heterogeneidad espacial multiescalar en los patrones migratorios.

Key Words:

• multiscale geographically weighted regression
• negative binomial
• spatial heterogeneity
• spatial interaction

关键词:

• 多尺度地理加权回归
• 负二项
• 空间异质性
• 空间交互

Palabras clave:

• binomial negativa
• heterogeneidad espacial
• interacción espacial
• regresión geográficamente ponderada a multiescala

Disclosure Statement

No potential conflict of interest was reported by the author.

Notes

1 Genetic algorithms (GAs) are often applied as an approach to solve global optimization problems. The initial bandwidths (initialization of the GA) will not affect the final results, but will affect the time to find the optimal bandwidths by affecting the convergence speed (Vlašić, Ðurasević, and Jakobović 2019). Although there are some more effective heuristic initialization methods, random initialization is the most commonly used.

2 The data and codes that support the findings of this study are available with the identifier at the private link https://figshare.com/s/0929779d36dd89aa0d38.

Additional information

Notes on contributors

Hanchen Yu

HANCHEN YU is a Visiting Assistant Professor in the Urban Governance and Design Thrust, Society Hub, The Hong Kong University of Science and Technology (Guangzhou), Guangzhou, China. E-mail: [email protected]. His research interests include spatial analysis, GIScience, spatial, spatiotemporal modeling, and spatial interaction modeling.