"[W]e tend to dwell usually on areas of high disease risk, often neglecting to note that areas with low rates may also give clues about disease causation."
p. 6:
"One virtue of GIS is that we are freed from the tyranny of census-based areal units: we might explore disease risk around a point source of pollution, or along a busy motorway instead."
p. 7:
Mention of kernel estimation, but not kriging (which we'll carry out in the geostatistics module).
"One of the key issues that arises in geographical epidemiology concerns the representation of exposure. Typically, this is represented by address at date of diagnosis. But this may be a far from perfect representation. How can GIS help us to represent exposures in other contexts (such as the workplace)? Can we model action and activity spaces, for example within a time-space framework suggested originally by Torsten Hagerstrand?"
p. 8:
The authors argue that "The added value of GIS is in the linking of databases...." But this is only half the value, in my mind: the powers of visualization (or "map comparison") are quite as valuable. One danger of visualization without analysis, however, is that it can result in the 'gee whiz' effect (p. 10).
p. 10:
"Jacquez suggests that we could do a lot worse than revisiting some of Karl Poppers's strictures on scientific explanation, setting up theories which one attempts constantly to refute or falsify."
p. 11:
"There is a need, within environmental epidemiology, to link environmental monitoring and modelling ... to a GIS so that health event data can be associated with the modelled outputs." [NB: Gatrell and Loytonen report that "The interpolation (kriging) approach results in an over-smoothed map of pollution levels and a regression approach yields better results." Kriging IS a regression approach! We'll see that kriging does not provide a single result, but rather a family, and relies on a prior modelling step which will effect the results. Did the author do a good job of modelling in this prior step?]
"[W]e seriously need, in geographical and environmental epidemiology, accurate models of exposure. We cannot always rely on residential address at the time of diagnosis if we wish to understand the processes giving rise to spatial patterninig of disease events."
p. 12:
"Loytonen provides an introduction to time geography for those to whom it is unfamiliar, and then considers recent and current progress being made in embedding time into GIS." Jacquez et al. at BioMedware are working on an STIS (Space-Time Information System).
Mention of "...contemporary visualisation research in epidemiology, seeking to move away from static, paper-based mortality atlas, towards electronic, interactive products with spatial statistical capability."