Graphics Reference
In-Depth Information
Adding Model or Statistical Information -
Overlaying (Statistical) Information
Guides may be drawn on a plot as a form of annotation and are useful for emphasiz-
ing particular issues, say which values are positive or negative. Sloping guides high-
light deviations from linearity. Fitted lines, for instance polynomial regression or
smoothers, may be superimposed on data not only to show the hypothesized overall
structure but also to highlight local variability and any lack of fit. Figure
.
plots the
timesfromthefirstandlaststagesofa
-kmroadrace.Alowesssmoother hasbeen
drawn. It suggests that there is a linear relationship for the faster runners and a flat
one for the slower ones.
When multiple measurements are available, it is standard practice in scientific
journals to plot point estimates with their corresponding confidence intervals. (
%
confidence intervals are most common, though it is wise to check precisely what has
been plotted.) Figure
.
displays the results of a study on the deterioration of a thin
plastic over time. Measurements could only be made by destructive testing, so all
measurements are of independent samples. he high variability at most of the time
points issurprising. Adjacent means have been joined by straight lines. A smoothing
function would be a better alternative, but such functions are not common for this
kind of plot. As the measurement timepoints are far apart and as there is only one
dataset, there is no overlapping here. hat can very oten be a serious problem.
Overlaying information, whether guides or annotation, can lead to overlapping
and cluttered displays. Good solutions are possible but may require individual ad-
justments dependingonthe shapeofthe data. Awell-spaced andinformative display
at one size may appear unsatisfactory and unclear when shrunk for publication.
Figure
.
.
Times for
runners for the last stage of a road race vs. their times for the first stage, with
a lowess smoother. Default scales from R have been used. Source: Everitt (
)
