dataviz/docs/07-check.md

202 lines
8.4 KiB
Markdown
Raw Normal View History

2021-04-10 22:10:25 +00:00
# CHECK Ensure visual integrity
2021-04-11 12:56:04 +00:00
CHECK covers all aspects of ensuring visual integrity in reports and
presentations.
2021-04-10 22:10:25 +00:00
_Ensuring visual integrity_ means that reports and presentations present
information in the most truthful and the most easily understood way by avoiding
misleading visuals.
2021-04-11 12:56:04 +00:00
This chapter covers avoiding manipulated axes and visualization elements, using
the same scales, and showing data adjustments.
2021-04-10 22:10:25 +00:00
1. [Avoid manipulated axes](#ch-1-avoid-manipulated-axes)
2. [Avoid manipulated visualization elements](#ch-2-avoid-manipulated-visualization-elements)
3. [Avoid misleading representations](#ch-3-avoid-misleading-representations)
4. [Use the same scales](#ch-4-use-the-same-scales)
5. [Show data adjustments](#ch-5-show-data-adjustments)
## CH 1 Avoid manipulated axes
Charts serve as a means to visually compare numerical values. Manipulated axes
defeat this purpose of explaining actual interrelations.
## CH 1.1 Avoid truncated axes
![Figure CH 1.1: Avoid truncated axes](img/ch-1.1.png)
2021-04-11 12:56:04 +00:00
Charts with value axes not starting at zero (“cut” axes) are not “wrong”in and
of themselves, but the message to be visually conveyed then does not correspond
to the numerical values upon which the chart is based. Therefore, value axes
should generally start at zero, see Figure CH 1.1.
2021-04-10 22:10:25 +00:00
2021-04-11 12:56:04 +00:00
One exception to this rule exists: charts with indexed data (e.g. if the value
for the index period is set to 100%) with only small variances from 100%. Here
“zooming in” on the variances could be of greater value than indicating the
absolute values (starting at zero). In this case, position the category labels
at the 100% line in order to avoid misinterpretations.
2021-04-10 22:10:25 +00:00
## CH 1.2 Avoid logarithmic axes
![Figure CH 1.2: Avoid logarithmic axes](img/ch-1.2.png)
2021-04-11 12:56:04 +00:00
_Avoid logarithmic scales_ because they do not allow the visual comparison of
values, see Figure CH 1.2. In business, very few applications for logarithmic
axes exist (e.g. comparing growth rates of different stocks in percent).
2021-04-10 22:10:25 +00:00
## CH 1.3 Avoid different class sizes
![Figure CH 1.3: Avoid different class sizes](img/ch-1.3.png)
2021-04-11 12:56:04 +00:00
If the categories represent ordered classes of elements (e.g. age classes) as
used for the visualization of distributions in histograms, use class sizes of
identical width (e.g. ten years). Otherwise, true visual comparability is
impossible, see Figure CH 1.3.
2021-04-10 22:10:25 +00:00
## CH 2 Avoid manipulated visualization elements
Displaying values differing by orders of magnitude can be challenging. Use
creative solutions instead of clipping visualization elements or cutting value
axes.
## CH 2.1 Avoid clipped visualization elements
![Figure CH 2.1: Avoid clipped visualization elements](img/ch-2.1.png)
2021-04-11 12:56:04 +00:00
Similar to “cut” axes, clipped visualization elements such as broken columns
make visual comparisons impossible, see Figure CH 2.1.
2021-04-10 22:10:25 +00:00
## CH 2.2 Use creative solutions for challenging scaling issues
2021-04-11 12:56:04 +00:00
![Figure CH 2.2: Use creative solutions for challenging scaling
issues](img/ch-2.2.png)
2021-04-10 22:10:25 +00:00
2021-04-11 12:56:04 +00:00
Creative visualization elements can be used to compare extreme values, e.g.,
displaying data in two-dimensional or even three-dimensional visualization
elements allows the comparison of values differing by orders of magnitude, see
Figure CH 2.2.
2021-04-10 22:10:25 +00:00
2021-04-11 12:56:04 +00:00
This rule must be clearly separated from the rules of section CH 3 “[Avoid
misleading representations](07-check.md#ch-3-avoid-misleading-representations)”
where area and volume visualizations are used improperly.
2021-04-10 22:10:25 +00:00
## CH 3 Avoid misleading representations
Graphical representations are misleading if the visual impression for the
observer differs from the underlying values.
## CH 3.1 Use correct area comparisons, prefer linear ones
2021-04-11 12:56:04 +00:00
![Figure CH 3.1: Use correct area comparisons, prefer linear
ones](img/ch-3.1.png)
2021-04-10 22:10:25 +00:00
2021-04-11 12:56:04 +00:00
Using two-dimensional representations (areas of circles, icons, or emblems) for
the visualization of data is only valid, if the size of these areas corresponds
to the underlying values. The visual perception will be misleading if the
diameters of circles or the heights of icons represent the values, see Figure CH
3.1.
2021-04-10 22:10:25 +00:00
## CH 3.2 Use correct volume comparisons, prefer linear ones
2021-04-11 12:56:04 +00:00
![Figure CH 3.2: Use correct volume comparisons, prefer linear
ones](img/ch-3.2.png)
2021-04-10 22:10:25 +00:00
Similar to areas, the visual perception will be misleading, if the
2021-04-11 12:56:04 +00:00
(one-dimensional) diameters or the (two-dimensional) areas of three-dimensional
visualization elements (spheres, cubes, etc.) represent the values, see Figure
CH 3.2. Even if their volumes represent the values, it is hard to perceive them
properly. Prefer linear comparisons instead.
2021-04-10 22:10:25 +00:00
## CH 3.3 Avoid misleading colored areas in maps
![Figure CH 3.3: Avoid misleading colored areas in maps](img/ch-3.3.png)
2021-04-11 12:56:04 +00:00
Different colored areas can be helpful to visualize the precipitation per square
meter or the population density. However, do not use colored areas for the
visualization of non-area-related figures such as market shares or return on
sales. Position columns or bars of identical scale in maps instead. By the way,
pie charts also work well here (an exception to the EX 2.1 “[Replace
pie...”](04-express.md#ex-21-replace-pie-and-ring-charts) because they can be
placed precisely at one point, like a city (see Figure CH 3.3).
2021-04-10 22:10:25 +00:00
## CH 4 Use the same scales
Proper visual comparison requires the usage of identical scales for identical
measure units, or if this is not possible a clear indication of the
difference. If possible, use a consistent scaling concept for the complete
report or presentation material.
## CH 4.1 Use identical scale for the same unit
![Figure CH 4.1: Use identical scale for the same unit](img/ch-4.1.png)
If presenting more than one chart of the same unit on one page, use the
2021-04-11 12:56:04 +00:00
identical scale for these charts, see Figure CH 4.1. In extreme situations
identical scales might not be desirable. In these exceptional cases the use of
scaling indicators (see [CH
4.3](07-check.md#ch-43-use-scaling-indicators-if-necessary) and [UN
5.2](09-unify.md#un-52-unify-scaling-indicators)) can be helpful.
2021-04-10 22:10:25 +00:00
## CH 4.2 Size charts to given data
![Figure CH 4.2: Size charts to given data](img/ch-4.2.png)
2021-04-11 12:56:04 +00:00
Using identical scales in multiple charts can be demanding if the values in the
charts differ by orders of magnitude. A good solution is adapting the chart size
to the given data, see Figure CH 4.2.
2021-04-10 22:10:25 +00:00
## CH 4.3 Use scaling indicators if necessary
![Figure CH 4.3: Use scaling indicators if necessary](img/ch-4.3.png)
2021-04-11 12:56:04 +00:00
There are several ways to overcome challenging scaling problems. _Scaling
indicators_, such as *scaling lines* and *scaling areas* indicating the same
numerical height (typically a power of 10) in all charts are helpful to assist
in comparing multiple charts (of the same unit) with different scales, see
Figure CH 4.3.
2021-04-10 22:10:25 +00:00
2021-04-11 12:56:04 +00:00
This guide suggests a _semantic design_ for scaling lines and scaling areas, see
UN 5.2 “[Unify scaling indicators](09-unify.md#un-52-unify-scaling-indicators)”.
2021-04-10 22:10:25 +00:00
## CH 4.4 Use outlier indicators if necessary
![Figure CH 4.4: Use outlier indicators if necessary](img/ch-4.4.png)
Certain values that are very big in comparison to other values are called
2021-04-11 12:56:04 +00:00
outliers. If an outlier is not important for business, e.g. a big relative
variance of a small value, then it is not appropriate to scale the whole chart
to this outlier. Therefore, use _outlier indicators_ for unimportant outliers,
see Figure CH 4.4.
2021-04-10 22:10:25 +00:00
2021-04-11 12:56:04 +00:00
This guide suggests a _semantic design_ for outlier indicators, see UN 5.3
“[Unify outlier indicators](09-unify.md#un-53-unify-outlier-indicators)”.
2021-04-10 22:10:25 +00:00
## CH 4.5 Use magnifying glasses
2021-04-11 12:56:04 +00:00
Another way to assist in scaling problems is to use “_magnifying glasses_” for
zooming in on a part of a chart with a bigger scale. Use an appropriate
visualization element to mark the part of a chart to be zoomed in and to link it
to a second chart displaying the zoomed part on a bigger scale.
2021-04-10 22:10:25 +00:00
## CH 5 Show data adjustments
For longer time series, currency and inflationary effects can bias the visual
impression, hiding the real development of business.
## CH 5.1 Show the impact of inflation
![Figure CH 5.1: Show the impact of inflation](img/ch-5.1.png)
2021-04-11 12:56:04 +00:00
Making inflation effects transparent helps avoid misinterpretations of time
series visualizations, see Figure CH 5.1.
2021-04-10 22:10:25 +00:00
## CH 5.2 Show the currency impact
![Figure CH 5.2: Show the currency impact](img/ch-5.2.png)
2021-04-11 12:56:04 +00:00
Similar to inflation effects, the adjustment of currency effects can help to
avoid misinterpretations, see Figure CH 5.2.
2021-04-10 22:10:25 +00:00
2021-04-11 12:56:04 +00:00
[← Increase information density](06-condense.md) | [Apply semantic notation
→](09-unify.md)