Chapter 4 Lecture
Uploaded by
John McaulayChapter 4 Lecture
Uploaded by
John McaulayWeek 5 - Class Agenda
1. Ch. 4 Lecture on Data Visualization
[Break: 10 min]
2. Introduction to Tableau and Lab 4-1 Preview
Chapter 4: Data
Visualization
Brandon Lock, Ph.D.
Baruch College, CUNY
Where we are now
2. Data
1. Data 3. Modeling 4.
Preparation
Analytics and Evaluation Visualization
and Cleaning
7. Key 8. Financial
5. The Modern 6. Audit
Performance Statement
Audit Analytics
Indicators Analytics
Identify the
questions
In the IMPACT cycle, Track Master the
we’re now going to
outcomes data
look at Communicating
Insights and Tracking Communicate Perform test
Outcomes.
insights plan
Address and
refine
results
Exhibit 1-1 The IMPACT Cycle
Learning Objectives
1. Determine the purpose of your data visualization
2. Choose the best visualization type for your dataset
3. Refine your visualization to communicate efficiently and effectively
4. Communicate your results in a written report
Discussion Question:
What are the characteristics of a good
data visualization?
Data Analytics are only as important and
effective as we can communicate and
make the data understandable.
What is the purpose of
your data visualization?
What type of
visualization should
Declarative you use?
Visualization
Q1: Is the information
conceptual or data-
driven?
Q2: Am I declaring
something or
exploring something?
Source: https://hbr.org/2016/06/visualizations-that-really-work
Is your visualization declarative or
exploratory?
Exploratory visualizations are used to gain Declarative visualizations are used to
initial insights into patterns within your explain results and present findings.
data.
Data-Driven Visualizations:
Qualitative vs Quantitative
Data
Qualitative versus Quantitative Data
Qualitative data is categorical data.
A. Nominal data is non-ordered.
(e.g. hair color)
B. Ordinal data can be ranked.
(e.g. gold, silver, bronze)
C. Textual data
(e.g. names, addresses, news articles, written documents)
Quantitative data is numeric
A. Interval data versus Ratio data
• Interval data shows data points over an ordered range where the difference
between two values is meaningful, but a zero value does not represent an
“absence of” the measure. (e.g. temperature (in Fahrenheit or Celsius), SAT
Score (200-800), credit score (300-850).
• Ratio data has the same characteristics as interval data but with a “true zero”.
When the variable equals 0.0, then there is none of that variable (e.g. employee
turnover rate, height, dollar amounts)
Quantitative data (continued)
B. Discrete data versus Continuous data
• Discrete data shows only whole numbers. (e.g. points in a basketball
game)
• Continuous data shows numbers with decimals. (e.g. distance)
C. Distributions describe the mean, median, and standard deviation of the data.
D. Proportions show parts of a whole (e.g. 55% cats, 45% dogs)
Q. Describe the type of data for each of
the following variables.
a. Instructor evaluations c. Names of companies listed on
(excellent, good, fair, weak, the Dow Jones Industrial
poor) Average
b. Weekly closing price of gold d. Market share of competitors
e. IQ test score
How do you choose the
right chart?
Common Visualizations for Qualitative Data
Making Comparisons:
Bar charts (Categories) Symbol maps
(Geographic Data)
Pie charts (Proportion)
Stacked bar chart (Bar + pie chart) Word clouds
(Text Data)
Tree maps (Compare hierarchical
values)
Heat maps (Compare many values)
Common Visualizations for Quantitative Data
Line charts Scatter plots (Relationship between 2 variables)
(Trends)
Box and whisker plots
(Outlier Detection) Filled map (Geographic Data)
Qualitative Data Quantitative Data
Here is a
Comparison: Outlier detection: summary guide
Bar chart
Pie chart
Box and whisker plot
of when to use
Stacked bar chart Relationship between different
Tree map two variables: visualizations.
Heat map Scatter plot
Geographic data: Trend over time: Also see
Symbol map Line chart
datavizcatalogue
Text data:
.com
Geographic data:
Word cloud Filled map
What’s wrong with this chart?
Inappropriate scales can introduce bias
Bad Visualization Better Visualization
How could we improve this visualization about
whose computer is attacked more?
If we care about individuals, an ordered
bar chart is a little more clear.
If we care about attacks by job function, a
bar chart can make comparisons more clear.
A stacked bar chart may be easier to
interpret (in less space).
Q. Which chart type is better for
working with dates over time, a bar
chart or line graph? Why?
DAA – pg. 153
Why is a line graph not intuitive here? What
would be a more appropriate visualization?
DAA – pg. 154
Better
Even Better (adding labels)
How can you refine your
charts?
Choose an appropriate scale and use colors
effectively.
How much data do you Would context or
need to show? What is the reference lines make the
baseline? 0 or something scale more meaningful?
else?
Do colors have meaning in
What do you do with your dataset? (e.g., green
outliers? (e.g., CEOs with for positive outcomes and
$1 salaries) red for negative
outcomes)
Communicating through
Writing
Keep these in mind for the final project write-up
Communication throughout the IMPACT model
Identify the questions: What is the Communicate insights: Explain your
objective of your project and why does visualization clearly with self-contained
it matter? descriptions. Describe the purpose and
take-away of your visualization.
Master the data: Describe your data
preparation process and data issues Track outcomes: Discuss what’s next in
you encountered in the ETL process. your analysis. What should be done
next to extend or further evaluate your
analysis?
Perform the test plan and address and
refine results: Describe your test plan,
results, refinements, interpretation,
and limitations.
Visualizations should be self-contained!
• Use variable names that are easy to interpret
• Use self-contained descriptions of your tables and figures (in a
caption or note) that allow readers to understand it without
reading your whole report.
Consider your audience and tone
Who is your audience and what do you want to convey to
them?
Craft different versions for different audiences.
Get feedback. Ask other people to read though your
writing to make sure you are clear.
Get to the point. Be clear, concise, and direct.
Summary
Key Questions to Answer When Developing a Data Visualization
• Determine the purpose of your visualization and type of data
• Is your purpose declarative or exploratory?
• Is the data being visualized quantitative or qualitative?
• Refining your visualizations
• Avoid bias and select the appropriate visualization for your data
• Make appropriate use of scale and color
• Communicate throughout the IMPACT Model
• Make self-contained visualizations
• Consider your audience and tone
Critiquing Visualizations
https://www.tableau.com/blog/viz-wiz-1-11