Geography Reference
In-Depth Information
b
Fig. 10 Crime setting mapped to degree grid squares covering Brazil (as % for each grid square):
a Thoroughfare; b Place of Residence; c Vehicle; d Commercial location; e Bank; f Pharmacy;
g Public transport station; h Public square; i Lottery; j Shop; k Petrol Station; l School; m Other
vehicle) the comments for the groups in Figs.
8
and
9
apply (coverage; emphasis
on the north east though vehicle-based crime is de-emphasised compared with the
equivalent state map; emphasis on the south).
Figure
11
maps each of the 22 identified reasons for crime incidents by degree
grid cell. The overall coverage and northeast/south emphasis comments for the
other groups apply here too, though strong patterns are less apparent here due to
the sheer amount of reasons leading to dilution of percentage magnitudes in each
of the maps.
In all groups of maps featured in Figs.
8
,
9
,
10
,
11
, the sparseness of data for the
less chosen attributes, only hinted at by the state maps, is starkly apparent (e.g.
Fig.
11
s representing police violence).
4 Results and Discussion
The approach taken in exploring the processed dataset with eXplorer was in two
parts. The first is the systematic mapping of each of the chosen attributes, to isolate
specific attributes that yield significant patterns. This is essentially the process
represented and described in
Sect. 3
, and ultimately, through identification of the
attributes with enough variable data to be interesting, effects a filtering of the
dataset. The second part of the approach was related to the linked exploration of
the dataset. This process roughly follows Shneiderman's (
1996
) visual information
seeking Mantra: overview, zoom/filter then details-on-demand. However, in this
case, the overview and filtering effectively occur at the same time. The in-depth
exploration was largely unstructured, save for two initial strategies, in effect a
starting point or trigger for visual analysis:
(a) Mapping any variable belonging to the crime setting group in the choropleth
display. This was adopted for reasons of cognitive affinity, that the attribute
most associated with the geography of crime (crime setting) has a locational
display method (the choropleth map)
(b) Separating any variables in the crime group from variables in the crime victim
group (principally that they would not be plotted against each other in the
scatterplot display), due to their semantic proximity (more will be said about
this later on).
However, these strategies mostly did not yield any notable discoveries, so the
following examples do not follow them rigorously.
Figure
12
shows some screenshots of the visual analysis on the state-based
displays. The choropleth maps in Fig.
12
have been classified into six classes
according to a natural breaks strategy. This is recommended for standalone display