Taxonomy of Data Visualization Methods - Data Visualization: A Successful Design Process - page 126

Description : Pie charts are probably the most contentious chart type and attract
much negative sentiment. While we know it is harder to accurately interpret angles
and judge the area of segments compared to other visual variables, the negativity
is arguably more a reflection of their relentless misuse. The inclusion of too many
categories and colors, 3D decoration, and poorly executed arrangement are often to
blame for this. Usually, a simple bar chart will suffice to demonstrate the part-to-whole
relationship. However, if you are determined to use a pie chart, always start the first
slice from the vertical position (to establish a sense of baseline), minimize the number
of categories being displayed (ideally maximum of three), and arrange the segments as
logically as possible. Variations include the donut chart, which is essentially the same
chart but with the center removed (to accommodate labels or nested donut charts).
Stacked bar chart (or stacked column chart)
Data variables : 2 x categorical, 1 x quantitative-ratio.
Visual variables : Length, color-hue, position, color-saturation/lightness.
Description : Stacked bars are fairly self-explanatory. They can be based on the
stacks of absolute values or standardized to show part of a whole breakdown, as
in following example. Colors and position differentiate the value categories. Where
the categorical values are ordinal in nature, it helps to sequence the values logically,
for example when you have sentiment data such as the Likert scale of disagree
(reds) through to agree (blues). This sequencing helps draw out the contrasting
composition of the sentiment from all categories. The only drawback of a stacked
chart is the difficulty in being able to accurate read bar lengths, as there is no
common baseline.