Graphics Reference
In-Depth Information
awkward. here are displays in scientific publications which are prepared by the au-
thorsandshouldbetheproductofcarefulandthoroughpreparation.Inthissituation
a graphic should match its context well. Whatever kind of data graphic is produced,
a number of general principles should be followed to ensure that the graphic is at
least correct.
Whether or not a graphic is then successful as a display depends on its subject,
on its context and on aesthetic considerations. It depends on what it is supposed
to show, on what form is chosen for it and on its audience. Readers familiar with
one kind of graphic will have no trouble interpreting another example of the same
kind. On the other hand, a graphic in a form which is new to readers may lead to
unanticipated interpretation di
culties. When someone has spent a long time on
a study and further time on the careful preparation of a graphic display to illustrate
theconclusions,theyareusuallyastonishedwhenothersdonotseewhattheycansee.
[his effect is, of course, not restricted to drawing graphics. Designers are frequently
shocked by how people initially misunderstand their products. How oten have you
stared at the shower in a strange hotel wondering howyou can get it to work without
its scalding or freezing you? Donald Norman's book (Norman,
) is filled with
excellent examples.]
Other factors have to be considered as well. A graphic may look different in print
than on a computer screen. Complex graphics may work successfully in scientific
articleswherethereadertakes timetofullyunderstandthem.heywillnotworkwell
asabriefiteminatelevisionnewsprogramme.Ontheotherhand,graphicswhichare
explainedbyacommentatoraredifferentfromgraphicsinprint.Ifgraphicsdisplayed
on the Web can be queried (as with some of the maps on http://www
.cancer.gov/
atlasplus/, discussed in Sect.
.
.
), then more information can be provided without
cluttering the display.
Scientiic Design Choices
in Data Visualization
2.4
Plotting a single variable should be fairly easy. he type of variable will influence the
typeof graphic chosen.For instance, histograms orboxplots are right for continuous
variables, while barcharts or piecharts are appropriate for categorical variables. In
both cases other choices are possible too. Whether the data should be transformed
or aggregated will depend on the distribution of the data and the goal of the graphic.
Scaling and captioning should be relatively straightforward, though they need to be
chosen with care.
It is a different matter with multivariate graphics, where even displaying the joint
distribution of two categorical variables is not simple. he main decision to be taken
for a multivariate graphic is the form of display, though the choice of variables and
their ordering are also important. In general a dependent variable should be plotted
last.Inascatterplot itistraditional toplotthedependentvariable onthevertical axis.
