Use this URL to cite or link to this record in EThOS:
Title: Simultaneous modelling of contextual and spatial interaction effects : towards a synthesis of spatial and multilevel modelling
Author: Dong, Guanpeng
ISNI:       0000 0004 5347 2052
Awarding Body: University of Bristol
Current Institution: University of Bristol
Date of Award: 2014
Availability of Full Text:
Access from EThOS:
This thesis explores suitable methodological frameworks for modelling geographical hierarchical data sets by bringing together spatial and multilevel modelling approaches. The primary reason is that for a geographically hierarchical data, two types of dependence effect are expected: the spatial dependence arising from the geographical proximity and the membership based group dependence effect. Three different frameworks are proposed in this study. First, in terms of extending the Geographically Weighted Regression (GWR) model to accommodate a hierarchical data structure, a contextualised GWR is developed. The contextual information is incorporated into a GWR model by adjusting the geographical weights matrix to measure proximity not only in terms of distance but also with respect to an attribute space defined by measures of each observation's neighbourhood. Next, by integrating spatial econometric and multilevel modelling approaches, this study develops a hierarchical spatial autoregressive model (HSAR) that allows for simultaneously modelling the group dependence effect and the spatial interaction effect at each level of the data hierarchy. A series of Monte Carlo simulation studies show that the HSAR model performs well in terms of retrieving the true model parameters accurately. In contrast, both the classic spatial econometric and multilevel models perform poorly in the presence of both spatial interactions and group dependence. Finally, turning to a usual two-level hierarchical survey data where individuals nest into areas, this research develops a spatial random slope multilevel model that can simultaneously model spatial interactions across the higher level areas and correlations between random effects within each area. The essence of the spatial random slope multilevel model is to regard random effect at the area level as a spatial process characterised by a multivariate conditional autoregressive (CAR) process. In combination, the thesis makes important contributions to a coming together of spatial and multilevel modelling for better investigating geographically hierarchical data.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available