Geography Reference
In-Depth Information
empirical Bayesian smoothing based on the first order queen contiguity. This means
that this time the smoothing is not based on the global mean value in the Czech
Republic but on the local mean in directly neighbouring municipal districts. This
map provides the most smoothed image of the situation and creates spatially
continuous areas comparing the previous two maps. Again, there are fewer areas
with the highest prevalence that are dissolved into other categories than in previous
cases.
Table
2
summarizes the statistical characteristics of the computed smoothed
prevalence rates. While the difference between the characteristics of raw and
globally smoothed data is not so evident, the difference between the locally
smoothed rate and the other two cases is easily visible in the case of mean and
standard deviation. On the other hand, the median and IQR confirm the usability of
the method, so any possible misinterpretation of the result is on a similar level as in
the other two cases.
Apart from the observed number of cases in municipal districts, we also have the
expected number of cases available because of the use of standardization and
Bayesian smoothing. Thanks to these two values, we were able to find the relative
risk (SMR) in administrative areas, which is in fact the ratio of the empirical
(observed) number of cases and the theoretical (expected) value expressed as a
percentage. This means that only one half of the expected cases were recorded in
areas where the relative risk is 50 %, which is why these areas are
than
they are supposed to be. Relative risk also allows a comparison of areas based on
riskiness (or vulnerability) with regard to the diseases.
The first relative risk choropleth map (Fig.
2
—top) depicts the relative risk,
when the expected values are based on population standardization in the adminis-
trative units without any application of Bayesian methods. One can see that the
eastern part of the Czech Republic (Moravia) is more affected than the western and
central part (Bohemia), and high-risk areas are located mainly in the north-east. The
globally smoothed choropleth map (Fig.
2
—middle) provides a completely differ-
ent image of the situation with an extremely smoothed surface, where only a few
districts deviate from the global average. Lastly, the locally smoothed values of
relative risk were computed and visualized in the map (Fig.
2
—bottom). The map
shows that Bohemia is riskier than Moravia and Silesia although the prevalence
maps tried to claim the completely opposite situation.
healthier
'
'
Fig. 1 (continued) infection cases per 1,000 people (
top
); Smoothed 5-year prevalence of cam-
pylobacter in population in municipal districts in the Czech Republic between 2008 and 2012,
which is obtained by global empirical Bayesian estimates of the prevalence rate based on binomial
distribution (
middle
); Smoothed 5-year prevalence of campylobacter in population in municipal
districts in the Czech Republic between 2008 and 2012, which is obtained by local empirical
Bayesian estimates of the prevalence rate based on the first order queen contiguity (
bottom
)
