Geoscience Reference
In-Depth Information
selection and layout of the data (methods that consist essentially of cleaning and
estimating the missing data); enhancement (by external data integration); reduction
(of dimensions analyzed by factorial analysis for example, or by a value estimated
by a statistical model). Furthermore, whereas the dependency between neighboring
observations in space and the heterogeneity of spatial processes involved made the
geographical data unfit for classical statistics methods, a number of specific methods
have been developed that integrate these characteristics and use them to give a
different insight to associations and spatial distributions [FOR 02, GUO 09,
MAT 06]. All these developments have led to making the specificity of geographical
data visible and to giving its own status to the spatial dimension taken into account
in analyses.
Nowadays, more than parallel approaches, the simple explorations, as well as the
explorations that we will qualify as abductive, to emphasize the going back and forth
approach mentioned above, and the confirmatory analyses, make up a set of
complementary approaches with sometimes very thin borders. Our position is at the
interface of these different approaches, combining different types of statistical
analyses to identify and describe the relationships embedded in the geographical
space and evolving over time. This could be the evolution of the correlation between
the density and the distance to the center within a metropolis, or the evolution of the
differences in the school performances of students according to gender, social
category and localization of attended school. It could also be the evolution of flows
of commuters between the municipalities of an urban area according to the evolution
of employment and the transport network. In all three cases, the purpose is to
understand a change in a relationship. The first case concerns the relationship
between variables characterizing spatial units, the second case concerns the
relationship between variables measured at the level of pupils and the third case
concerns the direct relationships between spatial units.
These approaches are conducted putting the reflection on the to-and-fro between
data types and types of analyses. This to-and-fro is facilitated on the technical level
by the development of new environments, but in the current context of
multiplication of information sources, there is a risk of shifting from “the
observation of the phenomenon to the observation of the data”
2
. Our choice is, on
the contrary, to favor an approach that gives a central place to the studied
phenomenon and to the entities associated with it. The question of the spatial and
temporal resolution of the observations that are linked to it benefit from being
processed at the conception stage of the objects of interest, as has already been
mentioned in Chapter 2; even if it is often raised at the time of the implementation of
analyses.
2 R. Laurini, EVS seminar February 17, 2014 Lyon: “
Are there fundamental principles in
Geographic Information Science?”
with N. Chrisman
.
