VISUALIZING YOUR
DATA EFFECTIVELY Kim Unger – Fall 2017
ABOUT ME
Senior Analytics Consultant, DataBrains
Former SSEF, ISEF, STS Finalist
Science Fair Judge (regional, SSEF, ISEF)
Data Visualization is my day-to-day job
kunger@[Link]
@WizardOfViz
AGENDA
Overview – four questions
Choosing the right chart/graph
Other visualizations
Visual Best Practice
Tableau – Visualization softwareFRE
E
FOUR QUESTIONS
1. What data is important to show?
2. What do I want to emphasize in the data?
3. What options do I have for displaying this data?
4. Which option is most effective in communicating
the data?
Time Part-To-
Ranking Deviation
Series Whole
Distributi Correlati Comparis
on on on
WHAT DO YOU WANT TO
SHOW WITH YOUR DATA?
TIME SERIES
VALUES DISPLAY HOW SOMETHING
CHANGED OVER TIME
Strip Plot (multiple)
Only when also
featuring
Bar Graph (vertical) Line Graph Dot Plot (vertical) distributions
To feature individual To feature overall When you do not
values and support trends and patterns have a value for
their comparisons. and support their every interval of
Quantitative scale comparisons time
must begin at zero.
Box Plot (vertical)
RANKING
VALUES ARE ORDERED BY SIZE
(DESCENDING OR ASCENDING)
Dot Plots
Bar Graphs
Quantitative scale must begin at zero
PART-TO-WHOLE
VALUES REPRESENT PARTS (RATIOS) OF A
WHOLE
Bar Graphs
Quantitative scale must begin at zero
WHAT ABOUT PIE CHARTS?
Commonly used to show parts of a whole
However…
Hard to judge relative size of pie slices – better at
differentiating length
Take up a lot of space to present little information
Require labels and good color contrast to even be usable
(often difficult)
Best use is when one overwhelmingly larger value
than the rest – no need to focus on actual values
DEVIATION
DIFFERENCE BETWEEN TWO SETS OF
VALUES
Bar Graphs Line Graph
Quantitative scale must being at zero Only when also featuring
time series or single
distribution
DISTRIBUTION
COUNT OF VALUES PER INTERVAL ALONG
QUANTITATIVE SCALE
Strip Plot (single)
When you want to
Bar Graphs Strip Plot
see each value
Quantitative Scale, must begin at (multiple)
zero When
comparing
multiple
distributions
AND you want
to see each
Line Graph Box Plots value
To feature overall When Comparing Multiple
shape of Distributions
distribution
CORRELATION
COMPARISON OF TWO PAIRED SETS OF
VALUES TO DETERMINE IF THERE IS A
RELATIONSHIP BETWEEN THEM
Scatter Plot
NORMAL COMPARISON
SIMPLE COMPARISON OF VALUES FOR A SET
OF ORDERED ITEMS
Dot Plots
Bar Graphs
Quantitative scale must begin at zero
Schematics
Explain how
Illustrations experiment was
conducted or OTHER
Flow Charts
design concepts
for engineering VISUALIZ
project
ATIONS
Raw data or
statistical summaries Tables A PICTURE IS
in well-organized
manner.
WORTH A
Convey important THOUSAND
details. Great to show WORDS
Photographs experimental setup,
or examples of
actual results
ADHERE TO DATA
PRESENTATION STANDARDS
IN YOUR FIELD
Judged by those often familiar with research field
Expected presentations of data in that field
Review scientific articles – how is data presented?
Are there graphs?
What kind?
What statistics are used?
Review schematics – are there specific icons?
Does the journal have a style guide?
Emphasiz
most important data
e
Orient graphs for legibility
VISUAL
BEST Organize graph/table
PRACTICE Avoid overloading graphs
S Limit
# of colors and
shapes
Inform through important text
DATA ANALYSIS VS DATA
VISUALIZATION
Traditionally enter data into spreadsheet (Excel)
Satisfactory, but strengths are in data analysis – not visualization
Time consuming to create graph variations
Alternative: Use data visualization software
FREE to students and teachers with .edu email
[Link]