-
Identifier
-
Thesis_MES_2019_CarmanL
-
Title
-
Text, Tables, Graphs, and Maps: Communicating with Data Visualization
-
Date
-
October 2019
-
Creator
-
Carman, Leslie
-
extracted text
-
Text, Tables, Graphs, and Maps:
Communicating with Data Visualization
by
Leslie Raine Carman
A Thesis
Submitted in partial fulfillment
of the requirements for the degree
Master of Environmental Studies
The Evergreen State College
October 2019
©2019 by Leslie Raine Carman. All rights reserved.
This Thesis for the Master of Environmental Studies Degree
by
Leslie Raine Carman
has been approved for
The Evergreen State College
by
________________________
Kathleen Saul, Ph.D.
Member of the Faculty
________________________
Date
ABSTRACT
Text, Tables, Graphs, and Maps:
Communicating with Data Visualization
Leslie Raine Carman
Scientists, researchers, and analysts visualize data that is then often shared with
the public. Knowing the audience for whom the visualization is intended for is important;
however, it can be difficult to predict how the audience will interpret these visualizations.
Elements of a visualization such as wording, font, size, color, theme, style, and the type
of visualization can affect what the audience interprets. To explore these elements of a
visualization, I created my own maps using GIS to depict an impactful topic, Cancer. I
surveyed an audience on the information they took away from those maps. I used this
survey to test these methods and to find out what my audience saw in my maps. My
survey results suggest that individual experiences affect which depictions people prefer
and how they interpret information, but despite these differences in experience, common
themes are pulled out by the audience.
Table of Contents
List of Figures .................................................................................................................................. v
List of Tables ................................................................................................................................. vii
Acknowledgements ....................................................................................................................... viii
Introduction ...................................................................................................................................... 1
Literature Review............................................................................................................................. 4
Poop Maps ................................................................................................................................... 4
Color and Color Vision Deficiency ........................................................................................... 11
Other Visual Considerations ...................................................................................................... 13
Themes and Styles ..................................................................................................................... 15
Conclusion ................................................................................................................................. 16
Methodology .................................................................................................................................. 17
Creating the Maps ...................................................................................................................... 18
Survey Methods-PreTest Phase ................................................................................................. 24
The Final Survey ........................................................................................................................ 28
Results and Discussion .................................................................................................................. 37
Demographics ............................................................................................................................ 38
Vision Deficiency ...................................................................................................................... 39
Illinois Knowledge and Residency ............................................................................................ 41
Maps........................................................................................................................................... 43
Preferred Map ............................................................................................................................ 51
Conclusion ..................................................................................................................................... 57
Future Study ............................................................................................................................... 57
Final Conclusions....................................................................................................................... 57
Bibliography .................................................................................................................................. 59
Appendices..................................................................................................................................... 63
Appendix 1 ................................................................................................................................. 63
Creating the maps ...................................................................................................................... 63
Appendix 2 ................................................................................................................................. 83
Qualtrics Survey......................................................................................................................... 83
Appendix 3 ................................................................................................................................. 84
Color Vision Deficiency Test .................................................................................................... 84
iv
List of Figures
Figure 1. ........................................................................................................................................... 6
Figure 2. ........................................................................................................................................... 6
Figure 3. ........................................................................................................................................... 7
Figure 4. ........................................................................................................................................... 8
Figure 5. ........................................................................................................................................... 9
Figure 6. ......................................................................................................................................... 11
Figure 7. ......................................................................................................................................... 14
Figure 8. ......................................................................................................................................... 16
Figure 9. ......................................................................................................................................... 19
Figure 10. ....................................................................................................................................... 20
Figure 11. ....................................................................................................................................... 21
Figure 12. ....................................................................................................................................... 22
Figure 13. ....................................................................................................................................... 23
Figure 14. ....................................................................................................................................... 24
Figure 15. ....................................................................................................................................... 25
Figure 16. ....................................................................................................................................... 26
Figure 17. ....................................................................................................................................... 27
Figure 18. ....................................................................................................................................... 29
Figure 19. ....................................................................................................................................... 30
Figure 20. ....................................................................................................................................... 30
Figure 21. ....................................................................................................................................... 31
Figure 22. ....................................................................................................................................... 32
Figure 23. ....................................................................................................................................... 33
Figure 24. ....................................................................................................................................... 34
Figure 25. ....................................................................................................................................... 35
Figure 26. ....................................................................................................................................... 36
Figure 27. ....................................................................................................................................... 37
Figure 28. ....................................................................................................................................... 38
Figure 29. ....................................................................................................................................... 40
Figure 30. ....................................................................................................................................... 42
Figure 31. ....................................................................................................................................... 43
Figure 32. ....................................................................................................................................... 46
Figure 33. ....................................................................................................................................... 48
Figure 34. ....................................................................................................................................... 49
Figure 35. ....................................................................................................................................... 51
Figure 36. ....................................................................................................................................... 53
Figure 37. ....................................................................................................................................... 55
Figure 38. ....................................................................................................................................... 63
Figure 39. ....................................................................................................................................... 65
Figure 40. ....................................................................................................................................... 68
v
Figure 41. ....................................................................................................................................... 69
Figure 42. ....................................................................................................................................... 71
Figure 43. ....................................................................................................................................... 73
Figure 44. ....................................................................................................................................... 74
Figure 45. ....................................................................................................................................... 75
Figure 46. ....................................................................................................................................... 76
Figure 47. ....................................................................................................................................... 76
Figure 48. ....................................................................................................................................... 77
Figure 49. ....................................................................................................................................... 78
Figure 50. ....................................................................................................................................... 79
Figure 51. ....................................................................................................................................... 80
Figure 52. ....................................................................................................................................... 81
Figure 53. ....................................................................................................................................... 82
Figure 54. ....................................................................................................................................... 84
Figure 55. ....................................................................................................................................... 85
Figure 56. ....................................................................................................................................... 86
Figure 57. ....................................................................................................................................... 87
Figure 58. ....................................................................................................................................... 88
Figure 59. ....................................................................................................................................... 89
Figure 60. ....................................................................................................................................... 90
vi
List of Tables
Table 1. .......................................................................................................................................... 66
Table 2. .......................................................................................................................................... 70
Table 3. .......................................................................................................................................... 70
Table 4. .......................................................................................................................................... 71
Table 5. .......................................................................................................................................... 72
vii
Acknowledgements
I would like to thank my thesis reader, Dr. Kathleen Saul, for being an amazing mentor,
an enthusiastic supporter and motivator, and a great friend. Kathleen has a level of
kindness and understanding that cannot be matched. She provided the support I needed
when in doubt and is always an encourager of unconventional ideas that allow her
students to grow and produce amazing work. I always looked forward to our weekly
meetings and appreciate all that Kathleen has done to make this thesis possible.
I thank my family for sharing their stories and love. My mom for exploring the country
with me on road trips and teaching me to write and communicate, my dad for telling me
stories and making me laugh, and my brother for living close to home to keep them busy
and for inspiring me with his own brand of weirdness.
I would like to thank my friends, Nicole and Averi, for sending me funny memes, going
out on adventures with me, and being there during life’s challenges. Thank you for all the
trips to the beach, trivia nights, pumpkin patches, sundaes on Sundays, and volleyball
leagues. 5 out of 7, would do again. Thank you Nicole for bringing me baby Purr-sula
and thank you Averi for coming with me to the shelter to bring home Hercules.
I would also like to thank Danny for cooking and cleaning during this last month of
writing, going on adventures, laughing at my memes, and spoiling my cats. Thank you
for your IT support and all your work to improve your cooking and tea brewing skills.
Hercules and Purr-sula enjoying the holidays.
viii
Introduction
Croaking frogs, clicking cicadas, clucking chickens, bellowing cows, and the
ballad of mourning doves woke me every summer morning in Illinois. I ran through
prairies and cornfields, watched monarch butterflies transform from caterpillar to
chrysalis, bottle fed goats, caught fireflies, and looked on as storms swirled the
landscape. I never thought I would leave such an enchanting place, to rip my roots from
the soil and migrate like the monarchs. These were the years before we had internet on
our farm, before social media and instant news.
As summer ended, we canned our tomatoes, put straw out for the chickens, and
put our gardens to bed. We would go back to school and decorate for Halloween. Living
in the middle of nowhere makes for boring trick-or-treating, so we would go to Aunt
Carol’s house every year as she lived in town. We dressed up as vampires, cats, JiffyPop, and batman, ate our yearly Halloween dinner at McDonald’s, and went to Aunt
Carol’s for candy. At least we did, until Aunt Carol was sick, and then gone. She had
breast cancer, something old people get, I thought.
We got older and Halloween gave way to Thanksgiving, which meant a visit to
my grandparents in Ohio where my dad had once been a kid. I looked forward to hearing
my grandma’s sweet West Virginian accent and my grandpa’s quirky Pittsburgh accent.
We baked pies and cookies with grandma, ate our Thanksgiving dinner, and went out
bowling with my grandpa. We carried out this tradition every year until my grandpa
couldn’t bowl anymore. The next year, we spent Thanksgiving at his funeral, cancer had
spread through his whole body. We barely aged and Thanksgiving turned to Christmas,
1
the time when we would spend the holidays with my mom’s family back in Illinois. We
opened presents, played games with my grandma, and watched NASA shows with my
grandpa, Richard. His last Christmas was his last day at home. Glioblastoma had come
suddenly and unexpectedly, and he was gone too soon. During his last few months I
became familiar with the process of chemotherapy and the frequent trips to the cancer
center. Richard inspired my interest in science and space, and later GIS and this thesis.
I left Illinois hoping to escape cancer. I continue to watch as my friends from
Illinois are touched by cancer. I see their updates on social media and see that they
experience the effects of cancer more often than my friends from other states. This thesis
is dedicated to my family and friends whose lives have been touched by cancer. To those
who have fought, are still fighting, have lost their fight, and have lost their loved ones.
Richard and me one Christmas Eve.
2
This thesis blends the science behind data, the art of visuals, and an understanding
of people. We create visuals to understand science and to convey that understanding to
others. Scientists, researchers, analysts, and data visualizers can influence what their
clients and the public understand about the data they are sharing depending on how they
craft these visualizations. Fonts, colors, and the order of the visuals are a few of the
important elements that can help or hinder how well an audience can interpret
visualizations. This thesis explores those elements and tests them using an online survey.
An anonymous survey allows for the discovery of what information is pulled from the
visualizations, what common themes and conclusions the audience comes to, and what
visualizations the audience finds most informative and appealing. This thesis builds on
the survey data with information from multiple disciplines to suggest best practices for
creating a visualization that is accessible, interesting, and informative.
3
Literature Review
Data can be presented in many ways, but how do we decide what the best way to
present that data is? We must think about who our audience is, how our data visualization
will be used and interpreted, and what mediums are best for our audience to understand
and interact with information. If we do not understand our audience, we cannot
effectively communicate the stories we aim to tell with the data. Elements in a
visualization such as color, font, themes, and styles all aid data visualizers in
communicating information and can help an audience interpret and engage in that
visualization. Each of these elements are unique to each visualization, but there are some
best practices that can be followed, as explained in the sections that follow.
Poop Maps
The visualizations below (Figures 1 - 5) show how a dataset can be visualized in
five different ways, two maps, two bar graphs, and a line graph. The first three
visualizations are all appropriately–or inappropriately–colored in brown. These data
visualizations tell the story of human feces left behind in the San Francisco area
effectively using color themes and scale to alert the viewer to the enormity of the issue,
although each in its own way. The first map produced by “Open the Books” (2019)
displays all incidences from 2011 to 2019 at the same time, showing the viewer that
uncollected waste is a huge and widespread problem. However, there are so many points
overlapping that location information, such as road names, cannot be read and the
distinction of land and water is not easy to see. Regardless, this visualization grabs the
4
viewer’s attention and elicits an emotional response. This is an effective method of
visualization to elicit this emotional response, but it does not tell the whole story.
While this first map shows the reported incidences all at once, creating a large
brown blob, the bar graph in Figure 2 breaks down the information through time, telling
viewers a different story. By including time, the bar graph sheds light on another
dimension of the issue: how incidences have increased from 2011 to 2018. The viewer
would not have been able to gather this information from the map visualization, but by
having an accompanying bar graph, the viewer can gather two different types of
information and two different stories from the same dataset.
The map from “RealtyHop” (2018) (Figure 3) takes this same dataset and cleans it
up (pun intended) to give us another dimension of understanding to the dataset, which is
an aggregate by location. By aggregating the points into polygons representing
neighborhoods, viewers can understand which neighborhoods are the messiest and should
be addressed first when tackling this issue. The “RealtyHop” map also provides
additional line and bar graphs that are multi-color instead of brown. These visualizations
do not follow color themes, but the line graph depicting incidents by year and state needs
these additional colors so that the viewer can separate the years from one another. The
bar graph visualized by “RealtyHop” displaying days of the week also displays multiple
colors. The different colors on the bar graph are not necessary for understanding the data,
and could be confusing to a viewer. The colors in the bar graph are the same as those in
the line graph but do not correspond to the same time frame of information. The brown
bar graph from “Open the Books” is arguably a better bar graph visualization than that
from “Realty Hop” due to this color issue.
5
Figure 1. Open the Books map showing all incidents from 2011-2019.
Figure 2. Open the Books bar graph separating the number of incidences by year by
2011-2018.
6
Figure 3. RealtyHop map showing incidents grouped by neighborhood in 2017.
7
Figure 4. Realty Hop line graph show incidents by year and month from 2011-2018.
8
Figure 5. Realty Hop bar graph showing incidents by the day of the week.
A map from “(Human) Wasteland” depicts incidents in a different time period and
displays the points as a poop emoji shown below in Figure 6. This map reduces some of
the clutter by only showing incidents from January 2015 (Blum 2017). This map may not
elicit the same emotional response as the “Open the Books” map but does allow the
viewer to better identify areas of the reported incidents. The map in Figure 6 also follows
the brown theme used by “Open the Books” and the “RealtyHop” map but takes the color
one step further by using poop emojis to display these incidents. It clearly cements the
theme of the map in the mind of the viewer. This map arguably displays the best use of
theme of all the visualizations.
9
The original author of this map, Jennifer Wong, also does something the previous
visualizations do not. She addresses the issues of other visualizations of this data and how
others have interpreted or used these visualizations (Wong, 2018). For example, she has a
marquee across the website addressing the misuse of her maps. Wong has not only stated
on her website, but in interviews as well, that these visualizations only depict incidents
reported through dialing 311, which are not necessarily reports of human waste. The
collected waste is not tested to determine if it is from people or animals (Eskenazi, 2018).
On her website, she also provides additional resources that address why there is an
increasing number of human waste incidents in San Francisco including the lack of
access to public toilets in the city (Swan, 2014). By providing these extra resources and
alerting first time viewers to her website about the issues of the data and its
representation, Jennifer Wong helps educate her viewers. Wong also uses text as an
element of her visualization. By doing so, she arguably provides the most information
within the map, and provides this information in an honest way. The effectiveness of
these visualizations can be explored further by looking into the elements behind
visualizations.
10
The map cannot be displayed from the (Human) Wasteland website as of October 2019.
Figure 6. The (Human) Wasteland map showing incidents in January 2015 (retrieved
from Blum 2017)
Color and Color Vision Deficiency
The way something appears to us is an important factor in the human experience
whether we are choosing a meal or viewing a representation of data. Color is an
important element of appearance and our vision as well as an important element of
creating a visualization. We must not only choose colors that are culturally and
thematically appropriate, but we must also consider how the colors we choose appear to
someone that views colors differently. Because color vision deficient individuals see the
world, and the visualizations they are presented, in a different way, we must create
visualizations with these individuals in mind.
For our audience members without color vision deficiencies, we can look to fields
of study such as consumer preferences and color theory when choosing a color palette for
11
data visualizations. For example, when choosing a salad, people tend to prefer salads with
colors that are saturated, have depth, and have high contrast such as reds and greens
together. Not only are high contrast colors preferred, but they also communicate higher
complexity to the viewer. (Paakki, 2019). These color preferences may have root in
biology (Palmer, 2010) and may be explained by color theory.
Color theory was originally proposed by Goethe who wrote the Theory of Color,
which organized color in orders and proposed the color wheel (Margo, 2018). Color
theory also suggests that when choosing color to represent a category or to flood a poster,
we must also know our audience. Factors such as culture (Cotgreave, 2019) or the colors
of an individual’s school or favorite football team can affect the colors the viewer prefers
(Palmer, 2010). Likewise, some colors may elicit unintended responses from your
audience (Field, 2018) and may be offensive (Brewer, 2016).
We must also match the data represented to the types of palettes: sequential
(ranging from light to dark), diverging (a midpoint color with extremes on either side),
qualitative (different hues), and bivariate (two sequential palettes) (Field, 2018). For
example, it would be more appropriate to use a diverging color palette over a sequential
color palette to visualize a political map that displays preference over two parties.
In addition, we need to consider the needs of individuals with color vision
deficiencies. Color vision deficiencies affect how hue is perceived by an individual which
means color palettes will be viewed differently by those individuals. By providing sets of
colors that can be discerned by those with color vision deficiencies, such as red and blue,
we can hope to reach a larger audience and allow our visualizations to be as accessible as
possible (Fields, 2018). If we are stuck with using palettes that are not color vision
12
deficient friendly due to organization constraints, such as red/yellow/green, we can mix
another hue such as blue into green, to make our visualizations more readable
(Cotgreave, 2019). Luckily, there are many tools data visualizers can access to create
color vision deficient friendly visuals such as Colorbrewer2 (2002) and tools to test those
visualizations such as Colblindor (2006). Textures (lines, dots, cross hatching) can also
be added in visualizations when color cannot (Wu 2018). Textures may also help those
without vision feel visualizations if the visualization is created in 3D using tools such as
3D printers (Field, 2018).
Other Visual Considerations
Although color is an important element in creating a visual, there are other visual
considerations that are just as important such as eye tracking, size, and font. Eye tracking
takes into account the way a viewer’s eyes move about a page, or in this case, a data
visualization. Elements such as size, location, and images affect where the eyes and
attention are drawn. Larger fonts will grab attention as well as images. The types of
images we select can also determine which images people tend to focus on. For example,
when viewing a visualization on workers compensation, viewers tended to focus most on
an image of the human body that appeared nude over images of maps and graphs
(Alberts, 2019).
Fonts can also be important in a visualization and improve readability if used
correctly. Lowercase letters tend to be easier to read while uppercase letters provide
distinction. Choosing a font with more space between characters may also be easier to
read, but may not always be possible to use due to space constraints (Field, 2018).
13
Depending on your audience, you may also want to choose a font that is easier for
individuals with dyslexia to read, such as Dyslexie font shown below in Figure 7. This
font helps those with dyslexia by using techniques such as letter angle, large letter
openings, and longer sticks (i.e. a lowercase h) to make each letter unique and harder to
mistake for another letter (Dyslexie Font, 2017). Fonts like these can make visualizations
more accessible to a wider audience but may not be accessible to all data visualizers due
to cost and restrictions an organization must follow when presenting papers, posters, or
other documents.
Figure 7. An example of Dyslexie font: slanted letters, larger character spaces, and
longer sticks give each letter unique characteristics so that they cannot be mistaken for
other letters as easily.
White space and the amount of information included are also important in a
visualization. Too much information can overwhelm your audience and they may not
view all aspects of your visualization (Alberts, 2019). Removing unnecessary information
can reduce clutter and the appearance of a busy visualization, allowing your audience to
gather important information without being overwhelmed or put-off. (Field, 2018).
Designing to a grid (a framework of crossing horizontal and vertical lines) can also help
organize information and reduce clutter (Cotgreave, 2019).
14
“So me think, why waste time say lot word when few word do trick?” -Kevin, The Office
Themes and Styles
Themes should also be employed when creating a visualization. Themes and
styles may be specific to an organization being represented or to the specific content
(Cotgreave, 2019). Aesthetic elements in a visualization such as font and color come
together to create a style that fits into the theme of the data you are visualizing. These
styles and themes can help elicit certain feelings from an audience (Field, 2018) and can
help your audience engage with your visualization (Nelson, 2018). For example, if you
were to map filming locations for Game of Thrones, then you may want to take styles
from the series (Houghton, 2019) as shown below in Figure 8.
15
Figure 8. A map depicting Game of Thrones filming locations using styles to create a
theme.
Conclusion
Color, placement of data, themes, the use of type, and white space all came
together in the visualization of the San Francisco poop problem presented at the start of
this section. While elements such as style help engage the viewer, other elements such as
font and color allow the viewer to read your visualizations. By improving the
accessibility of a visualization, you can allow your visualization to reach a wider
audience. Best practices that you are familiar should be used and researching your
audience and new techniques are important in creating better visualizations for each
project.
16
Methodology
Throughout my research, I was interested in what an audience understands from a
visualization. I wanted to know what my audience may see in a visualization that I don’t
and if there is any information that they are pulling from my visualization I did not
intend–similar to the (Human) Waste Project visualization that was mis-interpreted and
mis-used. To conduct this study, I created my own visualizations using GIS and chose a
topic to visualize that would be important to myself and others–cancer incident rates by
zip code in Illinois. To test the elements of this visualization, I provided the same
information four different ways and with two different color palettes. I wanted to test a
color vision deficient friendly color palette against a standard non-friendly palette. I was
interested not only in color vision deficient individuals’ preference, but also if non color
vision deficient individuals would prefer the more accessible palette. I created my
visualizations by following best practices such as creating to a grid, reducing clutter, and
following a theme. I created a survey to test these visualizations and first tested my maps
at the 2019 Esri User Conference. I used feedback and survey results from the conference
to improve my survey and send to my final test audience by using social media.
17
Creating the Maps
I created maps using ArcGIS 10.6.1 to visualize cancer incidence by zip code in
Illinois using data from the Illinois Department of Health. To select a color palette that
was appropriate for the dataset and color vision deficient friendly I used ColorBrewer2,
which provided many options. To select one of these options, I used a color palette
closest to the colors of stained cancer tissue when examined under a microscope. I
continued to use this color palette throughout my thesis including a poster and the
surveys I created in order to follow a theme. I created four maps using the same data, but
displayed by different classes and class groupings. I created a 15-class natural break
(jenks) map, an 8-class quantile map, a 5-class standard deviation map, and a 2-class
standard deviation map. I used the pink/purple color palette selected from ColorBrewer2
for the 15-class and 8-class maps shown below in Figure 9 and Figure 10, respectively.
For the 5-class standard deviation map, I used a diverging pink/green color palette
selected from ColorBrewer2 shown in Figure 11 and selected my own color palette of
purple/off-white for the 2-class standard deviation map shown in Figure 12. I also recolored the 15-class and 8-class maps with a standard Esri red/yellow/green palette to use
in my survey shown in Figure 13 and Figure 14, respectively. Step-by-step instructions
on how to create these maps are shown in Appendix 1.
18
Figure 9. The 15-class natural break (jenks) pink/purple map I created using ArcMap
10.6.1.
19
Figure 10. The 8-class quantile pink/purple map I created using ArcMap 10.6.1.
20
Figure 11. The 5-class standard deviation map I created using ArcMap 10.6.1.
21
Figure 12. The 2-class standard deviation map I created using ArcMap 10.6.1.
22
Figure 13. The 15-class natural break (jenks) red/yellow/green map I created using
ArcMap 10.6.1.
23
Figure 14. The 8-class quantile red/yellow/green map I created using ArcMap 10.6.1.
Survey Methods-PreTest Phase
After completing my maps, I created a poster to display at the 2019 Esri User
Conference in San Diego, California. To keep my thesis theme, I used the image of
24
invasive lobular carcinoma from John Hopkins Pathology as the background image and
used pinks and purples for text, boxes, and a custom QR code. I created the custom QR
code using QR Code Generator to select for a rounded style with a purple frame and pink
edges. I kept the poster simple with little text and kept the focus of the poster on the maps
and the QR code which linked to my survey. I included the 15-class natural break (jenks)
map, 8-class quantile map, 5-class standard deviation map, and 2-class standard deviation
map in the poster. I also included a call out text to highlight one of the zip codes with a
cancer rate above the standard deviation shown below in Figure 15.
Figure 15. The poster I created to display at the 2019 Esri User Conference in San
Diego, California
I had created business cards with the black QR code, shown in Figure 16 below,
to distribute at the conference to get more survey responses from respondents at other
25
conference events who may not have attended the poster session. The business cards also
made it easier for people to scan the code and complete the survey between or after
conference events.
Figure 16. Business cards I created with a custom QR code generated by QR Code
Generator to elicit survey responses from conference attendees.
I provided business cards in a pencil bag as well as paper versions of the test
survey in file folders next to my poster. The paper version of the survey was provided to
elicit survey responses from conference attendees who may not have or preferred not to
use mobile devices, shown in Figure 17 below. I also tested my poster’s readability for
color vision deficient individuals using Colblindor. The poster viewed through different
types of color vision deficiency are shown in Appendix 2.
26
Figure 17. My poster displayed at the 2019 Esri User Conference with paper copies of
the survey in the left file folder, an empty file folder to the right for completed surveys,
and a pencil bag with pencils to use for the paper surveys. The pencil bag also contained
extra business cards for respondents to take with to complete the survey when convenient.
To create both the test and final surveys, I used Qualtrics. I started the test survey
with demographic questions, then asked respondents about vision deficiencies, and their
history with and knowledge of Illinois. I then moved into the map-related questions: I
asked respondents what they learned from the 15-class natural break (jenks) map, the 8-
27
class quantile map, the 5-class standard deviation map, the 2-class standard deviation
map, the 15-class natural break (jenks) red/yellow/green map, and the 8-class quantile
red/yellow/green map. I was able to improve my survey in real-time based on the survey
responses and face-to-face conversations and completed a final survey improvement after
the conference. The final and improved survey questions were then saved as a new
survey in Qualtrics.
The Final Survey
To elicit surveys for the improved and final survey, I created public posts on
Facebook and LinkedIn asking respondents to take my survey and share the survey link
with others. The questions shown in Figures 18-27 below show all the questions that final
survey respondents could have received. Some questions had display logic and would
only display for respondents that selected certain answers. For example, the questions,
“Can you read this map? Please provide any details you would like to share.” would only
display if the respondent selected “Yes” to a previous question, “Do you have any type of
vision deficiency? If yes, can you please describe?”
28
Figure 18. Page 1 of the sent survey. By leaving the gender question as a text box
answer, the question is more inclusive. The final question, “Do you have any type of
vision deficiency?”, has display logic so that vision related questions are displayed to a
“Yes” answer later in the survey.
29
Figure 19. Page 2 of the sent survey. The final question, “Do you currently or have you
ever lived in Illinois?”, has display logic so that Illinois related questions on the next
page of the survey are displayed to a “Yes” answer. Respondents that selected “No” did
not see the next page of questions.
Figure 20. Page 3 of the sent survey. This page only displayed to respondents that
selected “Yes” to the previous question, “Do you currently or have you ever lived in
Illinois?”
30
Figure 21. Page 4 of the sent survey asking respondents about the 15-class natural break
(jenks) pink/purple map. The final question, “Can you read this map?”, only displayed to
respondents that selected “Yes” to the previous question “Do you have any type of vision
deficiency?”
31
Figure 22. Page 5 of the sent survey asking about the 8-class quantile pink/purple map.
The final question, “Can you read this map?”, only displayed to respondents that
selected “Yes” to the previous question, “Do you have any type of vision deficiency?”
32
Figure 23. Page 6 of the sent survey asking respondents about the 5-class standard
deviation map. The final question, “Can you read this map?”, only displayed to
respondents that selected “Yes” to the previous question, “Do you have any type of
vision deficiency?”
33
Figure 24. Page 7 of the sent survey asking respondents about the 2-class standard
deviation map. The final question, “Can you read this map?”, only displayed to
respondents that selected “Yes” to the previous question, “Do you have any type of
vision deficiency?”
34
Figure 25. Page 8 of the sent survey asking respondents to choose between the 15-class
natural break (jenks) pink/purple map and the 15-class natural break (jenks)
red/yellow/green map. I provided a smaller version with legends and a larger version
without legends for easier viewing in the Qualtrics survey format.
35
Figure 26. Page 9 of the sent survey asking respondents to chose between the 8-class
quantile pink/purple map and the 8-class quantile red/yellow/green map. I provided a
smaller version with legends and a larger version without legends for easier viewing in
the Qualtrics survey format.
36
Figure 27. Page 10 of the sent survey allowing the respondents to share any information
that was not covered with the survey questions.
I visualized and analyzed my survey results using Tableau which required
downloading the table from Qualtrics and preparing the table before uploading in
Tableau. Because Qualtrics is a more complex survey tool and may not widely be
available due to cost, I provided step-by-step instructions in Appendix 3 on how to clean
the exported Qualtrics survey results in Excel. Cleaning the exported results makes it
easier to use data visualization tools such as Tableau. Some institutions have a
Qualtrics/Tableau bridge in order to avoid this step, but this feature is not widely
available due to cost.
Results and Discussion
The following results are from the final sent survey sent to respondents through
Facebook and LinkedIn. Results from the survey used at the Esri conference are not
included. The results from the Esri survey were used to improve this final survey.
37
Demographics
Out of 61 responses, half of the survey respondents had no experience with
mapping using GIS or online tools. The respondents ranged in age from under 18 to the
65-74 age bracket with most respondents in the 25-34 age bracket (32.7%). All age
brackets up to age 74 were represented in the survey with no respondents aged over 74.
The question asking respondents about gender was a text box so that all spectrums of
gender would be captured. Of the respondents that reported gender, only female and male
and variations (such as f or m) were reported. Of these respondents, 79.6% were female
and 20.4% were male. These demographics were visualized using Tableau 2019.2 as
shown in Figure 28 below.
Figure 28. Survey respondent demographics displayed using Tableau 2019.2. Half of
respondents had no experience with GIS, most respondents were aged between 25 and
34, and most respondents were female.
38
Vision Deficiency
I asked respondents about vision deficiencies and asked respondents to describe
their vision deficiency. 45.9% of respondents reported a vision deficiency. Of the 28
respondents that reported a vision deficiency, 23 described their vision deficiency, which
included glasses and contacts, astigmatism, near-sightedness, far-sightedness, one
respondent with lazy eye, and one respondent with cataracts. I asked this vision question
hoping to have respondents with color vision deficiency so that I could see if there was a
difference between which map respondents preferred based on whether they had color
vision deficiency or not. The red/yellow/green maps would not be as easy to read to those
with color vision deficiency. However, no respondents that took the survey reported color
vision deficiency. It is likely that the ratio of female to male respondents contributed to
the lack of color vision deficient individuals that participated in the survey as men are
more likely to have color vision deficiencies than women. Color vision deficiency also
only affects 4% of the population (Field, 2018) and a larger group of respondents would
have increased the likelihood of some respondents with color vision deficiency. These
results are shown below in Figure 29.
39
Figure 29. Survey respondents’ vision deficiencies displayed using Tableau 2019.2. 28
respondents reported a vision deficiency and 23 of those respondents described their
deficiency. No respondents reported color vision deficiency.
40
Illinois Knowledge and Residency
Because the maps were of Illinois, I asked respondents about their knowledge of
Illinois and if they have ever lived in Illinois. Most respondents (86.9%) had at least
some knowledge of Illinois. Of the respondents that had not lived in Illinois, most had at
least some knowledge of Illinois with 58.1% having very little knowledge of Illinois. Of
the respondents that had lived in Illinois, most reported somewhat and quite a bit of
knowledge of Illinois, 43.3% and 46.7% respectively. Most respondents that had lived in
Illinois, had been there for 30 to 40 years (30% of respondents). These results are shown
below in Figure 30.
41
Figure 30. Survey respondents’ knowledge and residency in Illinois displayed using
Tableau 2019.2. Most respondents had at least some knowledge of Illinois with half of the
respondents residing in Illinois at some points in their lives.
I also asked respondents where in Illinois they had lived or currently were living.
The survey respondents were spread throughout Illinois. Many respondents gave multiple
answers with some respondents reporting city and others reporting county. Some
respondents gave vague answers such as “northwest”. These vague responses were not
42
visualized in Figure 31 below. In future studies, it may be beneficial to ask for specific
answers.
Figure 31. Two maps of where survey respondents currently live or had lived in Illinois
using Tableau 2019.2. Some respondents reported city while others reported county. The
county map also includes the city responses by displaying the county of the city that
respondents reported.
Maps
I asked respondents the same two to three questions about four different maps.
Respondents who selected “Yes” to the question, “Do you have any type of vision
43
deficiency?”, were prompted to answer three questions per map while the other
respondents replied to two.
1. What conclusions do you draw from this map?
2. Did anything surprise you? If yes, please describe.
3. (For those with a vision deficiency) Can you read this map? Please provide any
details you would like to share.
For the first map, the 15-class natural breaks (jenks) pink/purple map, 50 respondents
answered the first question, “What conclusions do you draw from this map?” Answers to
this question were similar, with a general sentiment that there were high rates of cancer,
yet some respondents saw patterns in the data while other saw no pattern at all. One
respondent commented that the map had a camouflage-like pattern. Three respondents
indicated that the patterns they saw indicated farming, rivers, and agrochemicals might
have caused higher incidences of cancer, resulting in their interest in buying organic
produce. It was interesting to see some respondents come to conclusions as to why there
may be higher rates of cancer. Examining the zip codes with the highest rates of cancer
indicated that rivers may be a factor as well as industry such as mining. Some
respondents sought more information on groupings such as income, population, rural vs.
urban, or age. Population had been accounted for in the analysis, but this information
was not shared with the respondents. I intentionally included as little information as
possible as to my methods to see what respondents saw in the information. Responses
such as this one indicate that when sharing information with the public it is best to
assume that some individuals will not understand all methods, and some text may be
44
needed to accompany a visualization in order to explain similar information. Many
respondents indicated that they noticed that urban areas had lower rates of cancer and
some indicated surprise at this observation. As the person creating these visualizations, I
was also initially surprised by this observation. Others commented that the size of the zip
code area might have affected the visual representation. This is another area that the
survey indicated text may benefit a visualization if it is to be shared with the public.
Health data such as cancer incidence is sensitive information and is often difficult to get
access to. When researching this topic, it was difficult to find data more granular than
county. Zip code data was the most granular data that I could get access to, and I could
only find zip code data that was publicly available for two states, Illinois and New York.
One respondent with no GIS experience even stated that it seemed that cancer rates
seemed to be tied to zip code. This response also indicates that additional text would be
beneficial as this was a map that examined cancer incidence by the reported zip code.
Overall, survey response attention focused on the darker purple areas with higher rates of
cancer.
Twenty-six respondents answered the second question, “Did anything surprise
you?”, for this map. The answers were similar to the first question, but mostly focused on
the urban vs. rural divide, with respondents stating that they saw lower incidence in urban
areas and higher incidence in rural areas. Fourteen respondents commented on the low
rates in the Chicago suburbs. I was also surprised by the low rates in the Chicago suburbs
when creating these maps.
Eight respondents answer the third question, “Can you read this map?”.
Respondents indicated that they could read the map with other respondents indicating that
45
they thought they could read it, but that the question made them doubt themselves. Those
who doubted themselves had no experience with GIS. This question was not intended to
make respondents doubt themselves but was intended for respondents with color vision
deficiencies. For future studies, it may be beneficial to have a check box of vision
deficiencies with color vision deficiency as an option so that this question only displays
for those individuals. For reference, this map is shown in Figure 32 below.
Figure 32. 15-class natural break (jenks) pink/purple map
For the second map, the 8-class quantile pink/purple map, 40 respondents
answered the first question, “What conclusions do you draw from this map?” The
answers were similar to the answers of the first map, with more respondents commenting
on the map pattern. A few of the respondents commented on the reduced number of
classes and the change in color. One respondent asked why the maps looked different
46
when they were said to show the same information, this respondent had no experience
with GIS. Another respondent with no GIS experience commented on specific cities and
areas in Illinois and guessed that wealthier, non-agricultural areas had no agricultural
pesticides and therefore lower rates of cancer. This survey respondent also guessed that
the darker areas may have chemical companies. When researching these dark areas,
chemical companies were present. This was the first time a chemical company was
discussed; for the first map, only agricultural practices had been mentioned.
Nineteen respondents answered the second question, “Did anything surprise
you?”, for this map. The answers to this question were similar to the questions for the
previous map although one respondent indicated that this 8-class map was easier to read.
This is not surprising as this map had fewer classes which is recommended in
cartography and was a quantile map, which creates a more visually interesting map.
Another respondent felt this map was trying to trick them with the change in number and
size of classes. Interestingly, this respondent had 1-5 years of experience with GIS.
Four respondents answered the third question, “Can you read this map?”. Two
respondents said yes, one was interested in land use and more information on agricultural
areas, and one wanted the population to be included in the map. Again, population
information had been used in the analysis, but this was not shared with the respondents.
For reference, this map is shown in Figure 33 below.
47
Figure 33. 8-class quantile pink/purple map.
For the third map, the 5-class standard deviation pink/purple map, 35 respondents
answered the first question, “What conclusions do you draw from this map?” The
answers to this question were similar to previous answers with some respondents
indicating that they did not understand the map. These respondents had no experience and
just 1-5 years of experience using GIS. I was surprised that respondents with some GIS
experience did not understand this map. I believe that the confusion from this map
stemmed from the standard deviation where some respondents may not have an
understanding of statistics or the language used in statistics. I intentionally did not share
text information about statistics for this map but these responses indicate that if maps like
these are to be shared with the public, then text information about the statistics would be
helpful. One respondent stated that statistics are not facts, this respondent had no
experience with GIS.
48
Eleven respondents answered the second question, “Did anything surprise you?”,
for this map. The answers to this question were similar to the previous questions in the
survey. One respondent stated that the cancer didn’t seem as bad with this map compared
to the other maps; this respondent had 1-5 years of experience using GIS. This response
is not surprising as this map has more white space than the previous maps.
Four respondents answered the third question, “Can you read this map?”, for this
map. Three respondents said yes; however, the last respondent felt that all of the maps
were only good for conversation and research starters. This respondent had no experience
with GIS. This respondent is not wrong, these maps are intended as a tool to research
areas with higher cancer rates. For reference, this map is shown in Figure 34 below.
Figure 34. 5-class standard deviation map.
For the fourth map, the 2-class standard deviation pink/purple map, 33
respondents answered the first question, “What conclusions do you draw from this map?”
49
The answers to this question were mostly similar to the previous questions in the survey.
More respondents were confused by this map as with the previous standard deviation
map and many respondents noted that the cancer rates seemed less severe. One
respondent suggested people reduce smoking and eat healthier foods. This respondent
had 1-5 years of experience using GIS. This map did not provide any data on health
habits of Illinois residents, but this topic could be explored further if researching why
these areas have higher cancer rates. Another respondent remarked that data could be
manipulated based on personal agendas. This respondent had no experience with GIS.
While the data in these maps was not changed, it can be visualized in different ways to
tell different stories, which was the purpose of the thesis. One respondent noted that there
was less detail and color in this map while others indicated that they did not like this
reduction in detail while others were able to pull out areas with cancer rates above the
standard deviation. This map is intended to identify the areas with the highest cancer
rates in order to research why these zip codes have higher rates of cancer. It is not the
most appealing map but is a great research tool.
Five respondents answered the second question, “Did anything surprise you?”, for
this map. The answers to this question were similar to the previous questions in the
survey.
Two respondents answered the third question, “Can you read this map?”, for this
map. Both said yes with one commenting that this map was their least favorite. This
respondent had no experience with GIS and likely would not be familiar with using GIS
as a tool for research. However, this map was not created to be the most appealing. The
8-class quantile map was intended for this purpose, so this response is not surprising. It is
50
also not surprising that fewer respondents answered the last questions. This indicates that
respondents were experiencing survey fatigue and shorter surveys may yield different
results. For reference, this map is shown in Figure 35 below.
Figure 35. 2-class standard deviation map.
Preferred Map
I asked respondents to choose between similar maps symbolized with different
color palettes. Thirty-six respondents answered this question for the first set of maps,
shown in Figure 36: the 15-class natural break (jenks) pink/purple and red/yellow/green
maps. Most respondents preferred the pink/purple map (55.6%) with under half of the
respondents preferring the red/yellow/green map (41.7%). Most respondents with no GIS
experience preferred the pink/purple map (58.8%), all respondents that had 5 to 10 and
more than 10 years of experience using GIS also preferred the pink/purple map. More
respondents with less than one year of experience using GIS preferred the
red/yellow/green map (54.5%) and respondents with 1 to 5 years of experience using GIS
51
were divided evenly between the pink/purple and red/yellow/green maps. This was
surprising initially as I expected respondents with no GIS experience to prefer the
red/yellow/green maps. The red/yellow/green palette is not an appropriate color palette
for sequential data, but those with a few years of GIS experience may prefer this palette
because they have not learned this yet and the red/yellow/green palette is standard when
learning GIS. These responses indicate that the red/yellow/green color palette preferences
are due to the teaching style of tools such as GIS and not due to a natural preference for
these colors. Younger respondents were more likely to prefer the red/yellow/green map
while older respondents were more likely to prefer the pink/purple map. Respondents that
were 35-44 largely preferred the red/yellow/green map over the pink/purple map
(85.7%). It is unclear why there may be this divide in age and color preference, but
shared events in time, such as childhood, may affect this preference. Both male and
female respondents preferred the pink/purple map over the red/ yellow/ green map with
54.2% of female respondents and 85.7% of male respondents. This was surprising as I
expected male respondents may reject a pink color palette. These results were visualized
using Tableau 2019.2 shown in Figure 36 below.
52
Figure 36. Survey respondents’ color palette preference displayed using Tableau 2019.2.
Most respondents preferred the pink/purple map.
Thirty-five respondents answered this question for the second set of maps, those
with the 8-class natural break (jenks) pink/purple and red/yellow/green maps. Most
53
respondents preferred the pink/purple map (57.1%) with under half of the respondents
preferring the red/yellow/green map (40%).
Half of the respondents with no GIS experience preferred the pink/purple map
while the only respondent with more than 10 years of experience using GIS preferring the
red/yellow/green map. I expect that this preference may be due to using red/yellow/green
palettes as a default in GIS. The other respondents with less than 1 year, 1 to 5 years, and
5 to 10 years of experience using GIS preferred the pink/purple map (60%, 66.7%, and
100%, respectively).
The only respondent that was under 18 preferred the red/yellow/green map and
most respondents in the 35-44 age bracket preferred the red/yellow/green map as well
(71.4%). Respondents in the 45-54 and 65-74 age brackets were divided on their
preference of maps. All other ages preferred the pink/purple map with 100% of those
aged 18-24, 58.3% aged 25-34, the only respondent aged 55-64. Both male and female
respondents preferred the pink/purple map over the red/yellow/green map with 54.2% of
female respondents and 57.1% of male respondents. One respondent had a strong reaction
to the red/yellow/green map stating that the map looked like “Digiorno Supreme Pizza
Barf”. I agreed with this assessment. These results were visualized using Tableau 2019.2
shown in Figure 37 below.
54
Figure 37. Survey respondents’ color palette preference displayed using Tableau 2019.2.
Most respondents preferred the pink/purple map.
Six respondents changed which map they preferred between the purple/pink
palette and the red/yellow/green palettes when choosing a preference in the 15-class then
55
8-class maps. Three respondents preferred the pink/purple color palette (map A) for the
15-class natural break (jenks) map but preferred the red/yellow/green palette (map B) for
the 8-class quantile map. Three respondents preferred the red/yellow/green color palette
(map B) for the 15-class natural break (jenks) map but preferred the pink/purple color
palette (map A) for the 8-class quantile map. This result was surprising as I expected
color preference to be the same for both maps.
56
Conclusion
Future Study
If I could conduct this study again, I would host a focus group to gain a more indepth understanding and detail from participants on what they understood from the maps,
what they liked or didn’t, and why. The focus group would remove the survey fatigue
element from the study and allow participants to build their responses from each other.
This could reduce confusion that some participants experienced in answering some of the
questions. For example, some participants misunderstood the question in my survey,
“Which of these maps do you prefer?” and stated that they preferred the maps with
legends. While these types of answers highlight areas where the survey could be cleaner,
they did not provide additional input as to which map color palette was preferable.
A future study could also provide an additional story map that draws out the
stories behind every zip code with cancer rates above the standard deviation, similar to
the callout I created in my poster (Figure 15). Many respondents did not like the 2-class
map but presenting this map as a story map–it’s intended purpose–may yield different
results. Respondents are also likely to have a different set of reactions to a story map than
those reactions elicited from the static maps.
Final Conclusions
An audience has a wide variety of knowledge and backgrounds but there are tools
and methods that we can employ to make visuals more understandable. While the first
57
step in the process is generally to know your audience, we must also understand that we
can never truly know everything about our audience. There are so many factors and
experiences that each individual participant or audience member brings to the table that
we cannot always account for. We ourselves are also people with biases and experiences
that shape our opinions and views of our audience and influences how we visualize the
data we share. These opinions we hold will never fully be true and so we must do our
best to create visuals to the best of our abilities. We must employ techniques that we are
familiar with and work diligently to continue to learn and grow as researchers and
visualizers to best serve our clients and the public.
The test of a truly exceptional data visualizer is not only determined by the
knowledge they have, but also by the knowledge they continue to grow. An excellent data
visualizer is open to new ideas, strives to learn, and unafraid to test these new ideas and
methods. As data visualizers, we must believe that perfection can never be reached, but
strive to come as close as possible with every visualization we create.
58
Bibliography
Alberts, A., & Cotgreave, A. (2019). Eye-tracking: what it teaches us about dashboard
design. Retrieved from https://www.tableau.com/learn/webinars/eye-trackingwhat-it-teaches-us-about-dashboard-design-1.
Blum, S. (2017, January 11). This Map Shows San Francisco Is Covered in Human Poop.
Retrieved from https://www.thrillist.com/news/nation/human-wasteland-mapplots-all-of-san-franciscos-poop.
Brewer, C., Harrower, M., & The Pennsylvania State University. (2002).
COLORBREWER 2.0. Retrieved from
http://colorbrewer2.org/#type=sequential&scheme=BuGn&n=3.
Brewer, C. A. (2016). Designing better maps: a guide for Gis users. Redlands, CA: Esri
Press.
Colblindor. (2006). Coblis - Color Blindness Simulator. Retrieved from
https://www.color-blindness.com/coblis-color-blindness-simulator/.
Cotgreave, A., Shaffer, J., & Wexler, S. (2019). Dashboards for insight and impact.
Retrieved from https://www.tableau.com/learn/webinars/dashboards-insight-andimpact-0.
59
Data Access and Dissemination Systems (DADS). (2010, October 5). American
FactFinder. Retrieved from
https://factfinder.census.gov/faces/nav/jsf/pages/searchresults.xhtml?refresh=t.
Dyslexie Font. (2017). Dyslexie font, the revolutionary font for people with dyslexia.
Retrieved from https://www.dyslexiefont.com/en/home/.
Eskenazi, J. (2018, October 15). 'Poop map' creator has a message: Stop crapping on San
Francisco. Retrieved from https://missionlocal.org/2018/10/poop-map-creatorhas-a-message-stop-crapping-on-san-francisco/.
Field, K. (2018). Cartography. Redlands, CA: Esri Press.
Harvard Library. (2018). Harvard Geospatial Library. Retrieved from
https://library.harvard.edu/services-tools/harvard-geospatial-library.
Houghton, J. (2019, April 12). On the Map: Game of Thrones filming locations.
Retrieved from https://communityhub.esriuk.com/geoxchange/2019/4/12/gameof-thrones-filming-locations.
Illinois Department of Public Health. (2019). Illinois State cancer registry. Retrieved
from http://www.idph.state.il.us/iscrstats/ZP/Show-ZP-Table.aspx.
60
Margo, C. E., & Harman, L. E. (2018). Helmholtz's critique of Goethe's Theory of Color:
more than meets the eye. Survey of Ophthalmology, 64(2), 241–247. doi:
10.1016/j.survophthal.2018.10.004
Nelson, J., & Esri Press. (2018). Mapping with style (Vol. 1). Retrieved from
https://www.esri.com/content/dam/esrisites/en-us/media/pdf/Mapping With Style,
Volume 1.pdf
Open the Books. (2019). 011-2019 San Francisco Human Waste Reportings . Retrieved
from https://www.openthebooks.com/map/?Map=32504&MapType=Pin.
Paakki, M., Sandell, M., & Hopia, A. (2019). Visual attractiveness depends on
colorfulness and color contrasts in mixed salads. Food Quality and
Preference, 76, 81–90. doi: 10.1016/j.foodqual.2019.04.004
Roshal, A. (2002). RARLAB. Retrieved from https://www.rarlab.com/download.htm.
Palmer, S. E., & Schloss, K. B. (2010). An ecological valence theory of human color
preference. Proceedings of the National Academy of Sciences, 107(19), 8877–
8882. doi: 10.1073/pnas.0906172107
61
QR Code Generator. (2019). Create your QR Code for free. Retrieved from
https://www.qr-code-generator.com/.
RealtyHop. (2018, September 29). Doo-Doo, the New Urban Crisis. Retrieved from
https://www.realtyhop.com/blog/doo-doo-the-new-urban-crisis/#figure1.
Swan, R. (2014, March 19). Occupied: San Francisco: Understanding the City Through
its Toilets. Yes, Really. Retrieved from
https://archives.sfweekly.com/sanfrancisco/occupied-san-franciscounderstanding-the-city-through-its-toilets-yesreally/Content?oid=2948873&showFullText=true.
Wong, J. (2018). (Human) Wasteland. Retrieved from
http://mochimachine.org/wasteland/#.
Wu, F.-G., Tseng, C.-Y., & Cheng, C.-M. (2018). The composition of visual texture
design on surface for color vision deficiency (CVD). Computers in Human
Behavior, 91, 84–96. doi: 10.1016/j.chb.2018.02.028
62
Appendices
Appendix 1
Creating the maps
1. I went to the Harvard Geospatial Library and selected “Launch HGL”:
https://library.harvard.edu/services-tools/harvard-geospatial-library
2. I then used the zoom tool and zoomed to Illinois to find the zip code tabulation
areas from 2000 as shown in Figure 38.
Figure 38. Zoom in to Illinois and scroll through the available data to find the
data titled “UA Census Zip Code Tabulation Areas, 2000 – Illinois” from the
U.S. Department of Commerce, Bureau of the Census, Geography Division.
3. I then added the UA Census Zip Code Tabulation Areas, 2000-Illinois from the
Harvard Geospatial Library data-set to my cart to download.
63
4. I downloaded the data-set, unzipped the file folder, and added the shapefile to
ArcMap 10.6.1.
5. I accessed the Illinois Department of Public Health website to download the zip
code data for cancer incidence from 1996 to 2000:
http://www.idph.state.il.us/about/epi/intropds.htm
The cancer incidence data is coded as shown in Figure 39 below and the
description of the data is as follows:
“The files include incidence data for invasive cancers only with the exception of
cancer of the bladder. Carcinoma in situ of the breast is provided in a separate
category. Non-melanoma skin cancers, cases reported with unknown or "other"
sex, and cases with an unknown age are omitted.” (Illinois Department of Public
Health, 2019).
64
FILE LAYOUT for ZIP code files
ZPCD9600.DAT
Record Format
(all fields are numeric)
Data Field
ZIP code
cancer site groups
age group code
sex code
diagnosis year group
Positions
1-5
6-7
8-8
9-9
10-13
Length
5
2
1
1
4
CODES FOR DATA FIELDS
ZIP code
Valid Illinois ZIP codes
cancer site group
1 oral cavity
2 colorectal
3 lung
4 breast - invasive (females)
5 cervix - invasive
6 prostate
7 urinary system
8 central nervous system
9 leukemias and lymphomas
10 all other cancers
11 breast in situ (females)
age group code
1 0-14
2 15-44
3 45-64
4 65 +
sex code
1 male
2 female
diagnosis year group
9600 diagnosed between 1996 and 2000
Figure 39. The file layout for Illinois cancer data by zip code tabulation
areas.
65
6. To download the data, I right clicked on the zip code download link
“ZPCD9600.EXE (871KB)”, selected “save link as”, and saved the .EXE file to
my hard drive.
7. The file could not be unzipped and extracted as usual, so I downloaded WinRAR
to extract the data. For my system, I downloaded “WinRAR x64 (64 bit) 5.80 beta
1”: https://www.rarlab.com/download.htm
8. I then right clicked on the .EXE file and selected “Extract Here”.
9. I then right clicked the ZPCD9600.DAT file that was extracted from the .EXE file
and selected “Open with > Notepad”.
10. I then right clicked in Notepad and selected “Select All” then copied and pasted
the text into an Excel workbook.
11. Each data entry displayed as a string of numbers in one column with the ZIP code
in positions 1-5, cancer site groups in positions 6-7, age group code in position 8,
sex code in position 9, and diagnosis year group in positions 10-13.
12. I used the following formulas in Excel to produce the results also shown in Table
1 below:
ZIPCode:
=LEFT(A2, 5)
CancerSiteCode:
=MID(A2, 6, 2)
AgeGroupCode:
=MID(A2, 8, 1)
SexGroupCode:
=MID(A2, 9, 1)
DiagnosisYearCode:
=MID(A2, 10, 4)
These formulas select certain characters in the string. The first formula for
ZIPCode selects the first 5 characters in the string and the other formulas select
the indicated number of characters in the string starting with the indicated position
number. For example, the second formula for CancerSiteCode selects two
characters starting at position 6 in the string.
Table 1. The first column labeled “Coded” shows an example of how the data displayed
after being extracted from the Illinois Department of Public Health website into Excel.
The remaining columns show the data after the Excel formulas were applied to each
column using the string in the first column.
Coded
ZIPCode
CancerSiteGroupCode
AgeGroupCode
SexGroupCode
DiganosisYearCode
60084 2419600
60084
2
4
1
9600
13. I then removed the “Coded” column and added the table in ArcMap.
66
14. I opened the table in ArcMap and used the ZIPCode column to create a new table
that shows the number of cancer incidences per ZIP code.
15. I joined the new table of cancer incidences per ZIP code to the Illinois ZIP code
tabulation area shapefile.
16. Because the dataset shows the total number of cancer incidences per ZIP code, I
downloaded Illinois 2000 population by ZIP code tabulation area from American
Fact Finder by filtering for topic > year > 2000 and then filtering for geographies
> 5-digit ZIP code tabulation area > Illinois > all 5-digit ZIP code tabulation areas
fully within/partially within Illinois:
https://factfinder.census.gov/faces/nav/jsf/pages/searchresults.xhtml?refresh=t
17. I used the “Profile of General Demographic Characteristics: 2000 Census 2000
Summary File 1 (SF 1) 100-Percent Data” table, the table ID is “DP-1”.
18. I removed extra population data from the table, removed characters incompatible
with ArcMap, and added the edited table in ArcMap.
19. I then combined the uploaded population table with the previously joined ZIP
code cancer incidence table in ArcMap.
20. I saved my shapefile with cancer incidences and population to my file
geodatabase, uploaded the new shapefile into my map, and removed the previous
shapefile that did not have this new information attached to it.
21. To account for cancer incidences per population in the ZIP code areas, I created a
new field in the attribute table of the shapefile and opened the editor in ArcMap.
22. In the new field, I calculated cancer incidence divided by population and
multiplied by 100 to show the percentage of cancer per zip code using the
following formula:
([Cnt_ZIPCode] / [HC01VC01]) *100
23. I created four maps classified four different ways. The first map was classified by
natural breaks (jenks) with 15 classes. The second map was classified by quantile
with eight classes.
24. The third map was classified by standard deviation with five classes and the
fourth map was also classified by standard deviation, but with two classes. The
two-class map was divided into cancer rates greater than 0.5 standard deviation
and those below 0.5 standard deviation.
67
25. To create a color palette that would be visible to those with color vision
deficiency, I opened www.colorbrewer2.org and selected sequential data,
colorblind safe, changed the number of data classes to the maximum of 9, and
then selected the pink/purple multi-hue. I chose the pink/purple multi-hue color
palette because it was colorblind safe, was a more visually interesting sequential
color palette, and was a color palette that is similar to cancer cell samples that
have been stained to study under the microscope as shown in Figure 40 below. By
choosing this color palette, I was able to create and follow a theme throughout my
thesis.
Figure 40. Histologic grade 1 invasive lobular carcinoma from John Hopkins Pathology
at https://pathology.jhu.edu/breast/types-of-breast-cancer/. This type of breast cancer is
the second most common type of breast cancer according to Johns Hopkins Medicine.
26. Since my first map had 15 classes, I needed to add additional colors to the color
palette generated by Colorbrewer2, which only had 9 classes. Colorbrewer2
defaults to display the hex code number so I selected RGB to display the RGB
values instead as shown below in Figure 41. To create the additional colors, I
selected new colors evenly spread between the existing colors in the color palette
provided by Colorbrewer2 and selected RGB values between the existing colors
as shown below in Table 2. For example, I created the color for values between
1.0% and 1.5%. The R value for that color, 254, is between the R value of the
previous R value, 255, and the next R value, 253.
68
Figure 41. Colorbrewer2 with arrows pointing to changing the number of data classes,
type of data, colorblind safe selection, and the RGB values. The color palette displayed is
the color palette I selected to create my 15-class natural break and 8-class quantile
maps.
69
Table 2. This table shows the RGB values used to visualize the 15-class natural break
(jenks) map with the resulting color shown below each value. The values highlighted in
pink were colors I created to add between the colors generated by Colorbrewer2.
15-Class Natural Break Map
<1.0%
1.0%-1.5%
1.5%-1.8%
1.8%-2.1%
2.1%-2.3%
R
255
254
253
252
252
G
247
235
224
215
197
B
243
233
221
213
192
2.3%-2.6%
2.6%-2.9%
R
251
250
247
221
200
G
175
159
104
52
30
B
187
181
161
151
140
4.4%-5.9%
5.9%-9.1%
9.1%-16.2%
16.2%-22.4%
22.4%-37.5%
174
150
122
100
73
R
2.9%-3.2%
3.2%-3.6%
3.6%-4.4%
G
1
1
1
1
0
B
126
120
119
115
106
27. To create a visually interesting 8-class quantile map, I used the 9-class sequential
colors generated by Colorbrewer2 after removing the middle color. The RGB
values used in this map are shown below in Table 3.
Table 3. This table shows the RGB values used to visualize the 8-class quantile map with
the resulting color shown below each value. This palette was created using the 9-class
sequential values generated by Colorbrewer2 with the middle color removed from the
palette so that the resulting map would be more visually interesting than the 8-class
sequential palette generated by Colorbrewer2.
8-Class Quantile Map
<1.8%
1.8%-2.1%
2.1%-2.4%
2.4%-2.7%
2.7%-2.9%
2.9%-3.2%
3.2%-3.6%
3.6%-37.5%
255
253
252
250
221
174
122
73
G
247
224
197
159
52
1
1
0
B
243
221
192
181
151
126
119
106
70
28. I used a different color palette generated from Colorbrewer2 for the 5-class
standard deviation map. To generate this color palette, I selected diverging data,
colorblind safe, five classes, and selected the pink/green diverging palette as
shown in Figure 42 below.
Figure 42. Colorbrewer2 with the pink/green diverging color palette. Arrows
point to changing the number of data classes, type of data, and the colorblind safe
selection. The color palette displayed is the color palette I selected to create my
5-class standard deviation map.
29. For the 5-class standard deviation map, I used the same color palette generated by
Colorbrewer2 without any changes as shown in Figure 42 above and Table 4
below.
Table 4. This table shows the RGB values used to visualize the 5-class standard deviation
map with the resulting color shown below each value. This palette was generated by
Colorbrewer2.
5-Class Standard Deviation Map
< -1.5 Std. Dev.
-1.5 – -0.50 Std. Dev.
-0.50 – 0.50 Std. Dev.
0.50 – 1.5 Std. Dev.
> 1.5 Std. Dev.
R
208
241
247
184
77
G
28
182
247
225
172
B
139
218
247
134
38
71
30. To create the color palette for the 2-class standard deviation map I used the
darkest purple from the 15-class natural break (jenks) map and created a purpletoned off-white color as shown in Table 5 below.
Table 5. This table shows the RGB values used to visualize the 2-class standard deviation
map with the resulting color shown below each value. I used the darkest purple
generated by Colorbrewer2 from the 15-class natural break (jenks) map for values above
the standard deviation and create a purple-toned off-white color for values below the
standard deviation.
2-Class Standard Deviation Map
< 0.50 Std. Dev.
>0.50 Std. Dev.
R
250
73
G
247
0
B
243
106
31. I also created two additional maps with a standard red/yellow/green palette. I used
the same 15-class natural break (jenks) map and 8-class quantile maps as before
but recolored them using the red/yellow/green palette generated by ArcMap
10.6.1 as shown in Figure 43 below.
72
Figure 43. The 15-class natural break (jenks) map recolored with a standard
red/yellow/green palette generated by ArcMap 10.6.1 with an arrow pointing to the
selected color palette.
32. To make the maps look clean and polished, I added a USA states layer from Esri
using the “Add Data” > “Add Data From ArcGIS Online…” tool and using the
search term “States” as shown in Figure 44 below.
73
Figure 44. ArcMap 10.6.1 with arrows pointing to the “Add Data” > “Add Data From
ArcGIS Online” tool and the selected USA States layer package managed by Esri.
33. I symbolized the USA States > USA States (over 1:3m) using black (RGB: 0, 0,
0) – which displayed as a dark gray on the map – with a white outline (RGB: 255,
255, 255).
34. I then added state name labels symbolized with white text (RGB: 255, 255, 255).
35. Next, I added a USA major cities layer from Esri using the “Add Data” > “Add
Data From ArcGIS Online…” tool and using the search term “USA Major Cities”
as shown in Figure 45 below.
74
Figure 45. ArcMap 10.6.1 with an arrow pointing to the selected USA Major Cities layer
package managed by Esri.
36. I then changed the symbology from the default population symbology to a single
symbol of a white circle size 8.00 with a black outline.
37. I then used the query builder under the definition query tab in layer properties to
only symbolize larger cities in and near Illinois so that map viewers could have a
reference to cities throughout the state. I symbolized Moline, Chicago, Galesburg,
Peoria, Normal, Champaign, Quincy, Springfield, and Carbondale, Illinois; and
St. Louis, Missouri, by using the following definition query which is also shown
in Figure 46 below:
"NAME" IN ( 'Moline' , 'Chicago' , 'Galesburg' , 'Peoria' , 'Normal' , 'Champaign' ,
'Quincy' , 'Springfield' , 'St. Louis' , 'Carbondale' ).
75
Figure 46. The definition query used in the query builder to select for cities in and near
Illinois.
38. To add the city names to the map, I went to the labels tab in layer properties by
checking the “Label features in this layer” box, selected “NAME” for the label
field, and symbolized with white test (Times New Roman size 12) with a black
mask sized 1.50 shown in Figure 47 below.
Figure 47. The symbology selected in ArcMap 10.6.1 to label cities.
76
39. The final maps that I created are shown below in Figures 48-53.
Figure 48. The 15-class natural break (jenks) pink/purple map I created using ArcMap
10.6.1.
77
Figure 49. The 8-class quantile pink/purple map I created using ArcMap 10.6.1.
78
Figure 50. The 5-class standard deviation map I created using ArcMap 10.6.1.
79
Figure 51. The 2-class standard deviation map I created using ArcMap 10.6.1.
80
Figure 52. The 15-class natural break (jenks) red/yellow/green map I created using
ArcMap 10.6.1.
81
Figure 53. The 8-class quantile red/yellow/green map I created using ArcMap 10.6.1.
82
Appendix 2
Qualtrics Survey
1. I exported my data from Qualtrics as a .CSV file and opened the file in Excel.
2. To make the Qualtrics download useable in Tableau, I edited the Excel sheet in a
separate tab and saved the file as an .XLSX file.
3. I deleted the Import ID row generated by Qualtrics (row 3) the first 14 Qualtrics
generated columns (A-Q). These columns contain information such as distribution
channel and time of survey completion.
4. I then changed the column headers with the question number and an abbreviated
version of the question. For example, question 1 “How many years of experience
do you have with mapping using GIS or online tools?” has the header “1 Years of
experience”. Some questions generate two columns, for example question 4 “Do
you have any type of vision deficiency? If yes, can you please describe?” allows
the survey respondent to select yes or no and provides a text area to the “yes”
selection. For these types of questions, both columns will have a header that starts
with the question number. For question 4, the columns are titled “4 Vision
deficiency” and “4 Text”.
5. I then removed the last column header row generated by Qualtrics (row 2) and
removed 4 survey responses that had not been completed.
6. Since question 4 was a text box question, I created a new column next to question
4 and entered the answers in the same format so that answers such as “f” and
“female” were all reported as “female”.
7. I also created a new column next to question 7 and entered the information as a
whole number since some answers were reported as text answers and reported
non-residence status as “0”.
8. I added 24 columns next to question 8 with the headers “8 City 1”, “8 County 1”,
“8 State 1”, etc. to enter the location by city, county, and state for each response
so that they could be mapped in Tableau. Answers that were vague, such as “West
central Illinois”, did not receive a city or county location in the newly added
columns.
9. I added two additional columns after questions 21 and 22 and entered “A”, “B”,
“No preference”, or “Null” so that the answers were all in the same format, as
most responses contained descriptive text.
83
Appendix 3
Color Vision Deficiency Test
To test the readability of the poster to different types of color vision deficiency, I
uploaded an image of the poster to the Colblindor Color Blindness Simulator and selected
for each type of color vision deficiency as shown in Figures 54-60 below:
https://www.color-blindness.com/coblis-color-blindness-simulator/
Figure 54. My poster viewed in the Colblindor Color Blindness Simulator as it
would be seen by someone with red-weak color blindness (protanomaly).
84
Figure 55. My poster viewed in the Colblindor Color Blindness Simulator as it
would be seen by someone with green-weak color blindness (deuteranomaly).
85
Figure 56. My poster viewed in the Colblindor Color Blindness Simulator as it
would be seen by someone with blue-weak color blindness (tritanomaly).
86
Figure 57. My poster viewed in the Colblindor Color Blindness Simulator as it
would be seen by someone with red-blind color blindness (protanopia).
87
Figure 58. My poster viewed in the Colblindor Color Blindness Simulator as it
would be seen by someone with green-blind color blindness (deuteranopia).
88
Figure 59. My poster viewed in the Colblindor Color Blindness Simulator as it
would be seen by someone with blue-blind color blindness (tritanopia).
89
Figure 60. My poster viewed in the Colblindor Color Blindness Simulator as it
would be seen by someone with blue cone monochromacy.
90
