lOMoARcPSD|63293293

Data Visualization Techniques: Identifying Misleading Charts

and Graphs
Statistics. (Montclair State University)

messages.pdf_cover_qr_code_label

messages.studocu_not_sponsored_or_endorsed_by_college
messages.downloaded_by
 lOMoARcPSD|63293293

Class 2

Data Visuals, Good

and Bad: Part I
by Amrit Parmar

messages.downloaded_by
 lOMoARcPSD|63293293

Introduction
In a world where data is so important, we all want to create clear and effective charts.
However, constructing accurate data visuals isnʼt something that is usually taught in school or during on-the-job
training. Instead, we often pick it up as we go, which means we often make choices that can result in misleading
visuals.
Similarly, incorrect data visualization has been frequently used in traditional media (especially TV and
newspaper) and social media to mislead the audience, whether intentionally or unintentionally.

Goals
In this chapter, we will learn about data visualization techniques, and call out BS on various misleading data
visualizations that have been published or presented in the media the years.

messages.downloaded_by
 lOMoARcPSD|63293293

Visualizing Categorical Data

For categorical data, we typically want to compare the frequency or proportion of categories.

Pie Charts Bar Charts Stacked Bar Charts

Great for showing the proportion Useful for comparing quantities Useful for comparing the
of categories within a whole across different categories composition of categories across
groups

messages.downloaded_by
 lOMoARcPSD|63293293

Visualizing Numerical Data

For numerical data, we typically want to show distributions, trends, or relationships

Line Charts Histograms Scatterplots

Ideal for showing trends over time Great for showing the distribution Useful for showing the
or continuous data of a single continuous numerical relationship between two
variable numerical variables

messages.downloaded_by
 lOMoARcPSD|63293293

Pie Charts
Pie charts are commonly used to represent categorical data, with each slice showing a category's contribution to
the whole. They work best with a small number of categories and for simple, straightforward visualizations.

Example: Let's visualize the distribution of pet types in a community by percentage, where the data shows 40% dogs,
30% cats, 15% fishes, 10% birds, and 5% others.

Distribution of Types of Pets Owned in a Community

Dogs Cats Fish Birds Others

messages.downloaded_by
 lOMoARcPSD|63293293

What do you think about this chart?

Point 1 Point 2 Point 3

Percentage of Americans who The individual slices of pie do Misleading data visualization
smoke marijuana in different not add up to 100%, this is the to convince the audiences that
years is a trend over time (i.e. most common way to mislead the percentage of Americans
quantitative data). A line the audience with a pie chart. who have smoked marijuana
graph should be used here has drastically increased
instead. since 1997.

messages.downloaded_by
 lOMoARcPSD|63293293

What do you think about this chart?

Point 1 Point 2 Point 3 Point 4

'Biggest COVID-19 The individual slices Even if we assume Thereʼs no mention

worries' is of pie do not add up that individual slices of the sample size
categorical data, to 100%, this is the added up to 100%, or source. Did they
and the author is most common way red, orange, and sample 10 people in
trying to show the to mislead the yellow are too your studio or 1000
proportion of audience with a pie similar as colors. Itʼs random Americans?
categories within a chart. always better to use What are we looking
whole. So a pie- three easily at?
chart is the correct distinguishable
graphical colors.
representation.

messages.downloaded_by
 lOMoARcPSD|63293293

Bar Charts: Best Practices

Bar charts are effective for comparing quantities across different categories.
They should have a vertical axis starting at 0 and consistent calibration.
Common mistakes include not starting the y-axis at 0, inconsistent scaling,
and improper ordering of categories.

1 Proper Axis Calibration

Always start the vertical axis at 0 for accurate comparisons

2 Consistent Scaling
Use uniform increments on both axes

3 Logical Ordering
Arrange categories in a meaningful order (e.g. chronological,
alphabetical, or by value)

4 Clear Labeling
Include clear labels and a legend if necessary

messages.downloaded_by
 lOMoARcPSD|63293293

Bar Chart Example

Average Monthly
Rainfall
This bar chart shows the
average monthly rainfall in mm
across major US Cities.

messages.downloaded_by
 lOMoARcPSD|63293293

What do you think about this graph?

Point 1 Point 2 Point 3

No calibration of vertical axis No consistency in calibration Clearly a misleading bar chart

(y-axis). A good bar chart of percentages on the vertical to forcefully show that NBC2
always has the vertical axis axis. 13% has the highest bar viewers are “not at allˮ
starting from 0. where in reality it should be concerned about the zika
the lowest. 28% doesnʼt even virus
have a bar to it

messages.downloaded_by
 lOMoARcPSD|63293293

What do you think about this graph?

Point 1 Point 2

A good bar chart always has the vertical axis Here the y-axis starts at 58% instead of 0%,
starting from 0 so that we have an even exaggerating the difference between the two bars.
comparison. Not starting the vertical axis at 0 is The difference between 77% and 65% is
the most common ways to mislead the audience significant, but the chart makes it look far more
and make a flawed bar chart. drastic than it actually is.

messages.downloaded_by
 lOMoARcPSD|63293293

How it should really look…

messages.downloaded_by
 lOMoARcPSD|63293293

What do you think about this graph?

Point

Here the y-axis starts at 155,000 instead of 0. This makes the change look much more drastic than it actually is.

messages.downloaded_by
 lOMoARcPSD|63293293

What do you think about this graph?

Point 1 Point 2

In this graph we can see that, x-axis is not Both axes should have consistent, uniform
organized chronologically. This is one of the most increments. For example, if the vertical axis starts
commonly used technique to make a deceptive with increments of 5 (0, 5, 10, 15, 20), it shouldn't
graph. It gives the false appearance of an overall suddenly jump to larger values like 70, 80, 90, 100.
decrease in cases across the board. In this case, the vertical axis is well-calibrated.

messages.downloaded_by
 lOMoARcPSD|63293293

From the March 12, 2020 broadcast of

The Rachel Maddow Show…

Point 1 Point 2 Point 3

COVID-19 cases by day is Take a look at the horizontal What effect do you think this
numerical data, and so a line axis (x-axis). It's jumping from spacing has on the
graph is the ideal graphical January 21 to February 21st to representation of the data?
representation to highlights February 28 and an increment What are the potential
trends over time. In this case, of 4 days. NEVER do this! motives of the ones who
you should be using a line There should always be an designed this graph?
graph instead of a bar chart. even spacing between the
dates on for correct data
visualization.

messages.downloaded_by
 lOMoARcPSD|63293293

Line Charts: Effective Use and Potential

Pitfalls
Line charts are ideal for displaying trends or changes over time. While it's generally acceptable not to start the y-axis at
0 for line charts, caution is needed to avoid exaggerating differences. Common errors include inconsistent axis
increments and improper scaling.

Choose Appropriate Data

1 Use line charts for continuous data or trends over time

Consider Y-Axis Start Point

2 Starting at 0 isn't always necessary, but be cautious about exaggerating differences

Use Consistent Increments

3 Ensure both axes have uniform, logical increments

Avoid Misleading Scales

4 Be wary of manipulated scales that distort the data's true nature

messages.downloaded_by
 lOMoARcPSD|63293293

Line Chart Example

messages.downloaded_by
 lOMoARcPSD|63293293

'Stand Your Ground', Effective Law?

Point 1

It looks like the gun death in Florida was on a rise

in 1990s and once the ‘Stand Your Groundʼ law
was enacted in 2005, then the murders by
firearm dropped drastically. Right?

Point 2

Wrong! The vertical axis is inverted! It doesnʼt

start at 0. What we would normally perceive as a
decreasing trend is actually an increasing trend,
and vice-versa.

messages.downloaded_by
 lOMoARcPSD|63293293

What Standing Your Ground really did

messages.downloaded_by
 lOMoARcPSD|63293293

What do you think about this report?

Point 1 Point 2 Point 3

Always start by asking However, look carefully at the It gives the graph a flattened
yourself: "Is the graphical y-axis and the uneven graph look, essentially making the
representation technique a intervals. What's the problem growth during the later dates
good one to represent the here? What effect does it seem less drastic and more
kind of data weʼre dealing have on the data? stabilized.
with?" COVID-19 Cases per
day is numerical data, so a
line graph is the ideal
graphical representation to
highlights trends over time in
this case.

messages.downloaded_by
 lOMoARcPSD|63293293

Conclusion: Spotting and

Avoiding Misleading
Visualizations
Misleading data visualizations come in many forms. To spot them, question
the source, examine the axes and scales, and consider if the visualization
method is appropriate. Practice and vigilance are key to identifying
deceptive graphs and charts.

Assess Appropriateness Examine Closely

Evaluate if the visualization method Look carefully at axes, scales, and
fits the data type data representation

Question Sources Think Critically

Consider the origin and potential Apply logical reasoning to interpret
biases of the data the presented information

messages.downloaded_by