Extending Geographic Profiling Likelihoods to Include a of Data Range Types Generated in Ecology and Epidemiology.
Abstract
This thesis revolves around the development of geographic pro ling, a spatial model originally
developed in criminology to identify areas that likely contain a suspect's home or workplace,
based upon where they committed their crimes. Geographic pro ling is still used to this day in
investigations of serial crime but has recently found a cornucopia of applications in ecology and
epidemiology. For example, the sightings of an invading species, responsible for the decline of
local wildlife, can be used to target their nesting sites for e cient removal from an environment.
Similarly, households testing positive for an infectious disease can be used to target breeding sites
of vectors responsible for the disease's transmission. Despite countless applications, geographic
pro ling models are limited to considering only a single type of spatial data; a set of points on a
map. The work I have conducted addresses this issue by specifying a set of geographic pro ling
models that can deal with multiple kinds of spatial data. In ecology, eld experiments surveying
alien species often record the location of an encounter and count the number of individuals present.
I make the rst development accordingly, by describing a model that produces interventions based
on spatial count data. I then describe the rst instance that geographic pro ling is applied to
simulated and real-world count data, comparing the conclusions of this new model to that of
an existing model that only considers the location recorded but not the number of individuals
counted. The next development focusses on applications in epidemiology. When testing members
of a household for an infectious disease, the test location and prevalence rate; the proportion of
individuals testing positive, are recorded. Hence, I build and test a geographic pro ling model that
draws conclusions via spatial prevalence data. In the nal part of the thesis, I return to analysing
the type of data common to geographic pro ling, a set of points on a map. I equip users with
the
exibility to specify varying assumptions about the process generating these spatial points,
ultimately leading to a model for better describing real data. To conclude, I summarise the impact
of each new development in this thesis and take a step back, establishing where geographic pro ling
ts within a universe of spatial models.
Authors
Stevens., Michael.Collections
- Theses [4116]