Use this URL to cite or link to this record in EThOS:
Title: Exploratory analysis of large spatial time series and network interaction data sets : house prices in England and Wales
Author: Cooper, Crispin H. V.
ISNI:       0000 0004 2750 4815
Awarding Body: Cardiff University
Current Institution: Cardiff University
Date of Award: 2010
Availability of Full Text:
Access from EThOS:
Access from Institution:
This thesis describes a combined exploratory analysis, on a fine spatial scale, of (i) England and Wales house prices, between the years 2000 and 2006; (ii) aggregate statistics taken from the UK census of 2001; and (iii) interaction statistics also taken from that census. The house price data is derived from individual transactions and analysed mainly in the form of ward level indices with a time resolution of 100 days. The study has twin aims: firstly, to improve understanding of the data set - which is large in nature - particularly with respect to exploring the interaction statistics; secondly, to improve the methods of exploratory analysis themselves. With respect to the aim of understanding the data, both migration and house price changes are visualised in a novel way, and regression is used to determine indicators of likely house price cross-correlations between different market areas. Ripple type effects are shown to be related both to reactive mechanisms, and to the composition of migration flows. Further visualisation shows that the market may be understood in terms of clusters with similar behaviour, or alternatively, in terms of market-driving and market-driven regions. Variables which can be used to define these clusters and regions are identified via further regression. With respect to improving the techniques of analysis, existing methods of visualising interaction data - based on clustering and linear ordering of points in geographic space - are extended to larger, hierarchical data sets and evaluated in this context. Novel approaches are presented for (i) construction of relative house price indices with minimal hedonic data, (ii) enhancement of time series predictions using cross-correlation data, and (iii) comparison of heterogeneous data sets via unification of all relevant information in the interaction domain, making it susceptible to analysis by regression aided with principal component based dimensionality reduction.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available