Publication type
Thesis/Degree/Other Honours
Author
Publication date
June 1, 2008
Abstract:
The investigation of associations between places, people and mental health is complicated and there are serious limitations in the current methodology. Using data from the Caerphilly Health and Social Needs Study (CHSNS), as well as the British Household Panel Survey, this thesis investigated some of these methodological issues.
Firstly, motivated by the skewed distribution of the Mental Health Inventory (MHI-5), methods for analysing the mental health score were examined. Five methods for deriving a cutpoint on the MHI-5 based on linking with the General Health Questionnaire, were investigated and cutpoints derived for each.
When investigating associations between place and health, hierarchical modelling is an extremely useful tool. Sparse levels of information are a potential problem when using this method. In the CHSNS, households represent a sparse level of context. A simulation study was conducted to explore the effect of sparse levels on the results of hierarchical analyses. It was found that, in general, the underestimation of fixed effect standard errors is smaller when a sparse level is included than when it is excluded.
Another methodological consideration for hierarchical modelling concerns the choice of geographical hierarchy to use. Administrative hierarchies have been criticised for being heterogeneous and arbitrary. An algorithm was developed to partition regions into areas that are homogenous with respect to a given set of variables, and was applied to Caerphilly county borough. The resulting sets of boundaries were compared with the 2001 census administrative boundaries. These new boundaries performed favourably in comparison with the administrative boundaries, indicating that administrative boundaries may not represent the most suitable hierarchy to employ in hierarchical analyses.
Subject
Notes
Index to Theses search
not held in Res Lib - bibliographic reference only
#512226