Geography Reference
In-Depth Information
the very limited range of web technologies we were allowed to use and the need to use only
types of maps and graphs which would be immediately understood by a very wide audience.
However, some technically much more interesting possibilities were outlined above and we
would be very interested in pursuing these collaboratively. Recent funding to extend the web
site comes with fewer technical restrictions, and we could add more interactive vector-based
visualizations using a combination of AJAX and Flash, or SVG.
The system also has some interesting analytic potential. Conventional statistical data li-
braries are very poorly suited to automated analysis because individual data values are spread
across many different tables of every size and shape, and metadata are held separately and de-
signed to be read only by humans. Conversely, our system holds both statistics and metadata
in absolutely consistent locations, making it easily accessible to analytic software such as the
Geographical Analysis Machine (Openshaw
et al.
, 1987). One interesting demonstration of
how easily accessible the data structure is to automated exploration is its performance with
Google, which is able to reach almost every page in the
Vision of Britain
web site: although
we exclude them from our usage statistics, a third of all page views are for Googlebots, but
in return the site features very strongly in search results.
One final caveat is necessary. The work reported here was very heavily influenced by data
archivists, and especially the work of the DDI Alliance. However, making the data standard
drive a visualization system required some extensions. Our ignoring the DDI concepts of
collection
and
study
, and introducing
database
and
theme
, are superficial. More importantly,
our system works only because we have introduced a detailed controlled vocabulary for
universes
, and because almost all nCubes in the system are defined as
additive
; as already
noted, the latest version of the DDI has adopted similar approaches. Both decisions com-
plicate creating the metadata, but our generation of derived values depends on our software
being able to identify which nCubes are built from related variables
and
share the same
universe. Even more importantly, virtually any published statistical table can be defined as
a non-additive nCube, but requiring additivity means that some individual census tables
have to be defined as four or five separate nCubes; however, almost all our repertoire of
visualizations are valid only for additive data.
The DDI Alliance is very much a creation of the data archiving community, but is in-
creasingly seeking to develop a 'life-cycle approach', as illustrated in Figure 13.7 (Miller
DDI: the life-cycle approach (
c
Figure 13.7
2005, DDI Alliance)
Search WWH ::
Custom Search