Graphics Reference
In-Depth Information
Previouslywepointedoutthattheuseofgraphicalrepresentationscouldbeastarting
point for data exploration. In other words, starting from these representations, the
main idea of graphical interactivity is to allow the user to visually query the data
(Unwin,
). hat is, once the data have been displayed on graphics, the next step
is to allow the user to manipulate the graphics in an interactive fashion, in order
to search for patterns in the data. We refer the interested reader to Wilhem's survey
on “the paradigm of the linked view”, which recently appeared in Rao et al. (
)
(Wilhelm,
).
According totheir basic aims, aconsistent taxonomy of graphical interactive tools
is as follows:
Finding patterns: clusters,outliers, unusual groups, and local densities are exam-
plesof features that should beinteractively inspected byusers.Very important is
the inspection of low-dimensional dependencies, which are helpful for dimen-
sion reduction.
Posing queries: ater the identification of interesting patterns, the user should be
able to highlight individual cases as well as subsets of data. he results of such
queries should be given in a graphic display.
Different interactive methods are used to carry out the previous tasks. he under-
lying idea that associates the different usable tools is the data consistency principle:
whateverchanges theusercauses,eachavailable data view(numerical andgraphical)
is consequently updated.
he analyst can thus act on the screen, generating parameter changes and causing
a new execution of the analysis and a subsequent redrawing of the graphical repre-
sentations. he user can take several different analysis pathways based on the use of
one or more interactive tools that allow the user to address different tasks:
Exploring the original dataset (drill-down functionalities);
Asking for a different representation of all the information (multiple views) or of
the information stored in a selected area of the graph (subset views);
Dropping points, units, and/or variables from the analysis (deletion): deleting
point(s) can help the user to evaluate the sensitivity of the analysis and of subse-
quent representation to particular units and in order to compare this sensitivity
among the different views;
Aggregating/disaggregating (grouping/ungrouping) units in order to reduce/in-
crease the level of detail of the representation;
Filtering the representation in ordertovisualize only part ofthe information. he
filtering criteria can be based both on values of one ormore variables (using suit-
able tools to build and/or query) and on indexes (the user can ask to graph units
characterized by given values of one or more indexes). With respect to the last
point, in particular, a set of statistical measures can be used to filter the relevant
information so as to reduce the amount of represented points according to the
importance of the information lying below;
Focusing on specific characteristics: the usercan decide whichfeature of the data
tovisualizeandthewaytodisplayit.hisincludesthechoiceofvariables,thescale
andaspectoftheplots,thepossibilitytoanimatetheplotsusingreal-timecontrols
