Geography Reference
In-Depth Information
elements identify, in lower case, the categories within the different variables; these elements
are linked by underscores. However, this is entirely to make them usable by the people build-
ing the system; the system itself would work just as well, and very slightly more efficiently,
if these strings were replaced by arbitrary ID numbers.
Secondly, the DDI standard says that every nCube must have a
universe
attribute
spec-
ifying the population it covers, and a
measurement unit
attribute, specifying whether the
numbers count people, or houses or hectares. One of our crucial changes was to re-define
both universes and measurement units as entities. This was necessary as our rules for gen-
erating derived values do not look just for nCubes with related variables: two nCubes might
both be based on two variables measuring sex and age, but it would be wrong to map from
an nCube classifying total population to one classifying deaths. We therefore also check
that the universes match, but this requires a fixed set of universe identifiers, not the simple
uncontrolled string of text specified by DDI 2.0. DDI 3.0 has similarly introduced controlled
vocabularies for universes.
,
13.3.3 Additional presentational elements
Our other extensions are designed to help organize a web site and assist in presentation. We
replace
studies
and
collections
, as high-level organizing concepts, with
database
and
themes
.
The
database
is the single entity in our particular system that everything else belongs to,
directly or indirectly, and is the starting point for browsing the structure.
Themes
are used to
divide up the statistical content and play a large role in the
Vision of Britain
web site, discussed
below. We defined 10 themes, based on an exhaustive analysis of the contents pages of
other people's social atlases: Population, Life and Death, Industry, Work and Poverty, Social
Structure, Housing, Learning and Language, Roots and Religion, Agriculture and Land Use,
and Political Life. All nCubes are assigned to a theme, and all variables are either directly
assigned to a theme or belong to a
variable group
which is assigned to a theme.
Our last new entity is the
rate
, which currently defines what maps can be created. A major
problem with earlier systems like the Great American History Machine was the ease with
which naıve users could create totally illogical maps (Miller and Modell, 1998). To some
extent, as discussed below, defining nCubes creates a set of ratios which are logically valid, but
they are not necessarily interesting; and some important ratios are between numbers from
very different structures, such as an infant mortality rate, which combines data on births
and deaths. Our
rates
are defined in terms of a numerator, a denominator and a multiplier,
the first two being cellRefs and the last a numeric constant. Even when the values come from
a single nCube, rate definitions record key aspects of conventional practice. For example, we
worry about the 'unemployment rate', defined as the number out of employment, divided
by the number economically active, multiplied by 100; not an 'employment rate, based on
the number in employment and multiplied by 1000'.
13.3.4 Implementing the DDS
Figure 13.2 is emphatically not a database structure diagram, as we further abstract the
DDI and hold all entities - everything in Figure 13.2 except the location map and the
Search WWH ::
Custom Search