Geography Reference
In-Depth Information
of the study area. To avoid biased results, it is necessary to take this factor into account
(see Bailey and Gatrell (1995) for a discussion about this issue). For example, if KE is
being conducted, and part of the kernel falls outside of the study area, then it will be
necessary to account for the area of the kernel that is not within the study area.
Applications and other issues
7.5
Many dif erent properties can be represented as point patterns. h ese include disease
events (e.g. Hill et al. , 2000), trees (e.g. Li and Zhang, 2007), and concentrations of
minerals in rock (Lloyd, 2006). Other applications are detailed by Bailey and Gatrell
(1995) and Diggle (2003). h is chapter provides only a brief outline of some key tech-
niques for the analysis of point patterns. Consideration of the population at risk (e.g.
accounting for the fact that disease rates tend to be higher in urban areas as there is a
greater density of people in such areas than elsewhere), edge ef ects, and testing of
point patterns (e.g. assessing how far a point pattern is clustered) are mentioned only
quite briel y. Further issues that are not covered include the extension of the analysis
of spatial point patterns to include a time element, analysis of marked point patterns
(i.e. points with values attached), and techniques for the identii cation of clusters (i.e.
specii c locations with clusters as opposed to clustering of point patterns in general).
Another issue that is not explored is the use of Monte Carlo randomization proce-
dures in assessing point patterns relative to complete spatial randomness. With such
approaches, multiple point patterns that are CSR are generated and these can be used
to assess signii cant departures of real point patterns from CSR. Bailey and Gatrell
(1995) describe such a procedure with respect to the L function. For the L function, it
is common practice to derive approximate upper and lower 5% coni dence intervals
from simulated values, and these values can then be plotted (see Fotheringham et al.
(2000) for an example). h e next section presents a case study that makes use of data
provided on the topic website.
Case study
7.6
h e following case study is based on locations in England that are represented on the
Gough Map of Great Britain (dating to c. 1360). h e data were captured as part of
a research project funded by the British Academy 1 (see Lloyd and Lilley, 2009).
h e aim of the project was to develop a digital representation of the Gough Map. h e
analysis entailed using kernel estimation to explore local variation in event (i.e. place)
intensity. Such an analysis is useful in that the results can be compared to those
obtained using other data sources (e.g. contemporary records of taxation, etc.) to help
1 http://www.qub.ac.uk/urban_mapping/gough_map/
Search WWH ::




Custom Search