Geography Reference
In-Depth Information
according to the category in which its corresponding attribute value falls (Waller
and Gotway
2004
). Although choropleth maps do not show continuously distrib-
uted values, they often portray densities (Rushton
2003
). Viewed in this way, one
can consider them as a visual tool for the analysis of spatial distribution of the
phenomenon.
All data were aggregated into municipal districts and counts were summarized
for all the years. Then these counts were standardized using the population data
from the Population and Housing Census of the Czech Republic. This step was
processed to allow a comparison of different municipal districts even with dissim-
ilar populations (number and age). Demographic characteristics of the population
as a whole were used as the basis for the indirect standardization. Based on the
standardized population and observed cases in administrative units, we were also
able to calculate the expected number of cases in a municipal district. The stan-
dardization itself serves for data smoothing. Furthermore, global empirical Bayes-
ian estimates of the number of cases based on binomial distribution is utilized, as
well as local empirical Bayesian estimates of the number of cases based on the first
order queen contiguity.
Statistical characteristics that summarize occurrence in the Czech Republic
together with standardized and smoothed counts are shown in Table
1
. The mean
and standard deviation of data does not provide the best estimation of the actual
state of the situation as the number of cases in a municipal district is usually closely
related to the population and is far from the normal distribution of the data.
Moreover, the disease has never been recorded in a significant number of districts.
That is why the median and interquartile range describe data better and all the tables
contain characteristics that were calculated for both the complete dataset and the
reduced dataset, which contained only municipal districts with recorded disease
cases.
The first figure (Fig.
1
—top) depicts the 5-year prevalence of campylobacter in
the population in municipal districts in the Czech Republic. It visualizes the number
of cases per 1,000 inhabitants in the area without any standardization or smoothing.
The map unintentionally emphasizes mainly areas with rather sparse populations
where the disease was recorded (dark areas) on the one hand, and areas with a high
number of people and recorded cases on the other. The next choropleth map
(Fig.
1
—middle part) presents the same original data but this time empirical
Bayesian smoothing was applied. The smoothing was based on the global mean
in the Czech Republic, which was used to smooth all the data. One can clearly
notice that even those areas without any occurrence of the disease now belong to the
category with lower prevalence. Also, more than 200 municipal districts that were
firstly marked as the highest prevalence areas were dissolved into categories with
lower prevalence rates. Finally, the third map (Fig.
1
—bottom) shows the smoothed
5-year prevalence of campylobacter, which is obtained by the application of local
