Data Visualization

Best Practices:
Informing Without Misleading

B AT C H : I O C L
Objectives

At the end of this module, you will be able to:

• Describe the meaning of "pre-attentive attributes" and "encoding."

• Describe the Gestalt principles of grouping.
• Describe best practices around the treatment of axes; color choices; and degree
of detail in a view.
• Choose the correct chart type for encoding quantitative message.
From Light to Thought

Resolution:
Objects 0.04mm wide (the width of a fine human hair) are just distinguishable by good eyes,
objects 0.02mm wide are not.
Detect about 635 separate points in 1 square inch area
 Limitations of Memory

If you want someone to make sense of the graph as a whole, then you must limit the
number of data components that encode distinct meanings to seven at most—and safer yet,
to no more than five
Pre-attentive Attributes &
Encoding
Pre-attentive Attributes

• Take a few seconds to count how many times the number "5" appears in
Figure 3
Pre-attentive Attributes

• Take a few seconds to count how many times the number "5" appears in
Figure 4
Pre-attentive Attributes
• Pre-attentive attributes are information we can process visually almost immediately
before sending the information to the attention-processing parts of our brain.
The advantages of visual analysis
The advantages of visual analysis

• We have limitations regarding the amount of information we can process

and pay attention to at the same time.
• Complex math or to complex textual information vs. visual attributes.

• Encoding is the method used for presenting data to the user.

• In a text table, text is used to encode the data values.
• Other ways to encode data: different mark types, color, size...
 Text table vs. line chart

• Humans find it more time-consuming to interpret text tables than to

interpret the same information in more visual ways.

Consider the effort required to answer questions with a text table, such as: How do our business
locations compare with each other? Did we gain or lose customers over a specific range of years?
Text table vs. line chart
Text table vs. bar chart

• It's not easy to scan this black-and-

white text table for the unprofitable
values.
Text table vs. bar chart

• In this version, color makes the

unprofitable values stand out more
visibly.
Text table vs. bar chart

• Using bar length as well as color to

encode the values makes them much
easier to compare.
• You could use sorting to control the
order.
Attributes aren’t perceived equally

• Some attributes are perceptually stronger than others.

Attributes aren’t perceived equally

• Some can be perceived

quantitatively and can therefore
be used to encode numeric
values, and others can't.
Attributes differ in their ability
• Only two of the pre-attentive attributes can be accurately used to encode quantitative
values:
• 2-D location (for example, the location of data points in a scatter plot)
• Line length (for example, the length of a bar in a bar graph).

• An object's size, as in its 2-D area (simultaneous perception of both length and width) -
we can tell that one object is bigger than another. However, it's difficult to determine by
how much they differ.

• Color intensity, can be quantitatively perceived to a degree—by making one value

darker, we can tell that it is greater than another—but not well enough to decode
specific shades into specific values without a lot of work.
Gestalt principles
What do you see?
• These principles address ways in which human visual perception automatically looks for and makes
assumptions about patterns and relationships between objects.

• Being familiar with them will help you design visualizations that viewers readily grasp.
Proximity

• When objects are in close

proximity, our minds naturally
infer a connection between
them.

• The lines in closer proximity to

each other may be interpreted
as related.
Proximity

• Arrange elements of your visualizations closer to each other if they are related:
• Titles should be placed near the charts they are related to.
• Color keys (legends) need to be located close to the charts they are used in.
• Filters/parameters should be positioned closer to the charts they influence.
• Charts related to each other, such as those representing the same metrics, should be
placed close to each other rather than to other charts.
Similarity

• Objects sharing the same color, shape, or size

are perceived as related or part of the same
group.

• For example, in a scatter plot that encodes

the marks with shape or color hue, viewers
will assume that the marks that look similar to
each other are related.
Similarity

• If the chart is merely colored without carrying any semantic meaning, it may be
harder to interpret than if left without color altogether.
• Use color for:
• Grouping to highlight similar characteristics. E.g., assigning color to a scatterplot can
convey additional characteristics of the elements.
• Directing the audience’s attention to elements you consider significant, serving as a focus
mechanism within your visualization.
Enclosure

• This principle, suggests that objects ‘enclosed’

within a defined area belong to a group.
• People will perceive a relationship between objects
that are enclosed by a boundary.

• In this example, enclosure draws attention to the

area within the boundaries.
Enclosure
Continuity

• People assume a relationship between

objects that appear continuous, such as
objects that are lined up in a row.

• The human eye follows lines, curves, or a

sequence of shapes in order to determine a
relationship between design elements.

• These marks showing earthquake locations

are not actually connected, though they seem
to be continuous.
Continuity

• In charts, this principle primarily involves

sorting and order.
• Bar charts are more comprehensible when
arranged from larger to smaller, time charts
from past to future, and so forth. This
organization allows us to perceive them as
one continuous whole.

• Captions, legends, and other visual elements

should be organized in conjunction with the
chart — consider alignment, indentation, etc
Connection

• If objects are connected, we perceive

them as a unified entity.
• This principle holds more influence than
common colors and shapes.

• People will assume a relationship exists

between connected items, such as values
that are connected by a line chart.
Connection

• This principle is especially evident in

networks and line graphs — thanks to the
lines, we understand that the dots are
interconnected, leading us to infer that
they relate to the same thing or share
similar characteristics.
Displaying and configuring
axes
Truncating axes: Bar charts

• The truncated axis exaggerates

the difference between Europe
and the other regions.

• Because people expect bars to

represent entire values, it is
likely that this truncated chart
would mislead people.
Truncating axes: Line charts

• For example, line charts are generally

expected to show change rather than entire
absolute values.
• Sometimes, truncating the axis of a line
chart helps to make the change more
apparent, as in the example.

• Just be aware that you can make the

amount of change look dramatically bigger
or smaller depending on where you start
the axis.
Standard axis orientation

• Viewers will assume that a chart's axes conform to the standard orientation they
are accustomed to.
Independent and dependent variables

• When two variables are included in

a visualization, viewers will expect to
see the independent variable on the
X axis and the dependent variable
on the Y axis.
Dual Axis Chart

• When a dual axis chart includes the same units on each axis, synchronize the axes to
improve consistency and clarity.
• Note, however, that when each axis contains clearly different units, such as age and
temperature, synchronizing the axes is neither necessary nor useful.
Expectations about color
Cultural and natural assumptions

• Color assumptions will vary depending on cultural background. For example,

Western cultures typically use red to signify a hazard, problem, or loss.
• However, some Asian cultures consider red to symbolize luck, celebration, and
long life.
Cultural and natural assumptions
 Unintentional color relationships

Consider the use of color across an entire visualization, and synchronize the meanings of the
colors used across the views in a dashboard or story.
Unintentional color relationships

• In this chart, viewers may

believe the dark and light
green bars are related, but
those colors were
assigned based on
alphabetical order and
weren't meant to suggest
a relationship.
Choosing colors
Choosing colors
How much detail is too
much?
How much detail is too much?
How much detail is too much?

• A best practice is to give viewers

access to more breadth and depth
of information without showing it all
at the same time, so that they can
investigate the data behind the
overall insight themselves.

• Dashboards provides a general

visual overview that sequences the
information so users can interact
with the view to see more detail
Use the 5 second test
• If users can’t figure out your viz within 5 seconds, rethink your approach.

Here are some ways to pass the test:

• Position the most important view goes on top or top-left
• Position legends near their views
• Avoid using multiple color schemes on a single dashboard
• Use 5 or fewer views in dashboards
• Provide interactivity
• Carefully word Titles, Axes, and Units
• Highlight key facts and figures
Choosing the Right
Chart Type
Scales of Measurement

The scale of measurement determines the amount of information contained

in the data and indicates the most appropriate data summarization and
statistical analyses.

Nominal Ordinal Interval Ratio

AMOUNT OF INFORMATION
Types of Variables

• To determine the types of pre-processing needed and algorithms to be used

• Numeric
• Continuous
• Discrete (Integer)
• Categorical
• Ordered (low, medium, high)
• Unordered (male, female, other)
• Binary (Flag) (Yes, No)

56
Primary types of quantitative messages &
How to graph them

1. Nominal Comparison

2. Time series
3. Ranking
4. Part-to-whole

5. Deviation
6. Frequency distribution
7. Correlation.
 Why not pie diagram?
If the area of the small circle below equals a value of 1, what is the area of
the large circle?

Our eyes are great at comparing differences in 2-D location and differences in line length,
but not 2-D areas and angles.
We are pretty bad at reading angles

• Look at the pie chart below and try to place the slices in order from largest to smallest.
We are pretty bad at reading angles

• Look at how easy it is to compare the percentages using the bar graph below,
which displays the same values:
• This information is much easier to read when presented in a table than it was when awkwardly
arranged around the periphery of the pie. So why use a graph at all?

• Why show a picture of the data if the picture can’t be decoded and doesn’t present the
information more meaningfully?
Try to follow the changes of these various companies and how they compare to one
another through time.

Notice how easily you can do it, however, using the bar graphs
People love dressing up their pie charts
today to look mouthwatering

Please avoid doing this

**Be mindful that correlation doesn’t guarantee a relationship.
Graph Selection
Matrix
Thank You !