A unified statistical modelling approach to deal with space, time and scale effects underlying "Old" and "New" geographical data
Statistical modelling tools that can simultaneously and properly deal with space, time and scale effects of geographical data are rather limited. Yet, they are in great demand especially with emerging new forms of data with fine spatio-temporal resolution, such as the GPS and cellphone data, being increasingly exploited to understand human activities and their interactions with geographical environment.
The study introduces two novel statistical models: 1) A multi-level temporal autoregressive model to deal with potential space, time and scale effects underlying GPS trajectory data; 2) A locally adaptive multi-level spatial econometric model to estimate scale effects, spatial dependency as well as the spatial connection structure of units.
The developed methodologies are applied to explore geographical contextual effects on human daily activities and build neighbourhood quality index with property transaction data. The R code for implementing the models has been made publicly available and is very simple to use.
Dr. DONG Guanpeng, University of Bristol
10:30am July 6, 2018
Room 1501, IGSNRR
Key Laboratory of Regional Sustainable Development Modeling of CAS