57
SAMPLING AND VARIANCE ANALYSIS IN REGGIO CALABRIA
Vincenzo Barrile* and Rossella Nocera**
* DIMET University of Reggio Calabria - Italy
Phone: ++39 965 875200, Fax:++39 965 875247, e-mail: barrile@ns . inq. unirc . it
** IGP ETH Zurich - Switzerland
Phone: ++41 1 6333049, Fax: ++41 1 6331101, e-mail: nocera@qeod. ethz . ch
ISPRS Commission VI, Working Group 3
KEY WORDS: Spatial analysis, Non - linear regression, Variance analysis, Geomarketing.
ABSTRACT
In this paper, it is presented the way used to examine data coming from a GIS which has been built, in order to be an
useful instrument to people who work in economical and social fields.
The monitored territory is the city of Reggio Calabria. Firstly the acquisition of data has been made: digitizing of the
map, collecting data from many different archives, surveying, etc. Then the GIS has been built up, finally some
examples of data analysis have been presented.
Starting from the implemented GIS, it’s possible to obtain data or attributes, corresponding to some, previous defined,
entities. A statistical analysis of those data has been done, in order to study the relations among them and to allow for
doing some models, able to represent the real situation and to show the real trend.
1. THE PRINCIPAL STEPS TO BUILD UP
THE STATISTICAL MODEL
The implemented GIS allows to know better the territory,
giving geographic information, integrated and merged,
with many other types of information (statistical, detected
and results from market studies).
In the present case the most important types of data are:
• data which describe the physical aspect of the
territory;
• data which describe social and economical aspects.
All the collected data, the GIS technology and the
statistical analysis allow for creating a model which
represents the real situation and shows the real trend.
The statistical analysis have, at its heart, a model which
attempts to describe the structures or relationships, in
some objects or phenomena on which measurements (the
data) are taken.
The process of developing a statistical model varies
depending on whether a classical hypothesis - driven
approach (confirmatory data analysis) or a more modem,
data - driven approach (exploratory data analysis) is
followed.
In many data analysis projects, both approaches are
frequently used. In classical regression analysis, the
residuals are usually examined, by using of exploratory
data analytic methods, for verifying whether underlying
assumptions of the model hold. The goal of either
approach is a model which imitates, as closely as
possible, in as simple a way as possible, the properties of
the objects or phenomena being modelled. Creating a
model usually involves the following steps:
1. Determine the variables to observe (preliminary data
modeling). In a study involving a classical modeling
approach, these variables correspond to the
hypothesis being tested. For data - driven modeling,
these variables are the link to the phenomena being
modeled.
2. Collect and record the data observations (data
acquisition).
3. Study graphics and summaries of the collected data
to discover and remove mistakes and to reveal low
dimensional relationships between variables
(qualitative robustness).