Graphics Reference
In-Depth Information
is the connection between MDA techniques and graphic interaction tools (Klinke,
). he sotware was designed as a data-mining tool to work with huge amounts
of data.
Data presented in this section come fromthe Italian Survey of HouseholdIncome
and Wealth (SHIW), a large-scale household survey run biannually by the Bank of
Italyonapanelofabout
Italianhouseholds(
individuals)spanningabout
Italian municipalities (Bank of Italy,
). he SHIW began in the
s with
the aim of gathering data on the income and savings of Italian households, and it
is widely used in studies on saving behavior by Italian households. It collects de-
tailed information on Italian household demographic characteristics, consumption,
income,and balance sheet items. Forour analysis, weconsider data of the last survey
(year
).Foranexhaustive description ofthedataandrelatedissues,seeD'Alessio
et al. (
).
For the sake of simplicity, among the available indicators we have selected are the
following:
ETA
:age(years)
Q
: working status (
=employed,
=self-employed,
=unemployed)
AREA
:geographicalarea
(
=northeast,
=northwest,
=central,
=south,
=islands)
YL
: compensation of employees
YT
: pensions and net transfers
YM
: net income from self-employment
YC
:propertyincome
CD
: consumption (durables)
CN
: consumption (nondurables)
S
: i g
AR
: real assets
AF
: financial assets
PF
: financial liabilities
BD
:consumerdurables
Data exploration and feature extraction can be well presented through the visual ap-
proach. Unfortunately, static views presented in this section do not allow us to ap-
preciate the human-machine interaction analysis power that comes from MDA.
hefollowing resultsrefertoatwo-step analysis (factorialanalysis andclustering).
Looking at the point configuration on the factorial plans, the user can modify the
analysis parameters or add and remove variables or statistical units.
VITAMIN-S sotware was designed to be used also by users not having a high
levelofskill inMDA.Indeed,specifyingthetypeofvariables andthedesiredanalysis
strategy is enough to start.
InthisexampleweclassedallvariablesascontinuousexceptQandAREA
.Inpar-
ticular we used the AGE variable and the two categorical variables as supplementary.
hen we specified factorial analysis plus clustering as a strategy. Once these parame-
terswereestablished, VITAMIN-Sperformsthewholeanalysis andthentheusercan
act on the graphic displays. Any action performed on one display causes methods to