online paper
- Kuhn, M. (2019, March 27). The caret package. Github. https://topepo.github.io/caret/
Outline for Assignment 1
- Refer to the Documenting Research Guide for assistance in organizing your research and developing your outline.
- You can find this guide in the Useful Information folder, as well.
Using the research guide and the assignment 1 instructions, develop your outline. Submit the outline in an MS Word document file type. Utilize the standards in APA 7 for all citations or references in the outline. Ensure that the document includes your name. Do not include your student identification number. You may use the cover page from the student paper template, but it is not required.
The assignment 1 instructions are at the bottom of this content folder.Submit your outline on or before the due date.By submitting this paper, you agree:
(1) that you are submitting your paper to be used and stored as part of the SafeAssign™ services in accordance with the Blackboard Privacy Policy;
(2) that your institution may use your paper in accordance with UC’s policies; and
(3) that the use of SafeAssign will be without recourse against Blackboard Inc. and its affiliates.
Documenting
Research Guide Last Revised: 12/27/2020 1
Contents
……………………………………………………………………………………………………………….. 2
Outline Example coinciding with Unit 3 …………………………………………………………………………………………………. 4
………………………………………………………………………………………………………………………………………… 7
Example Research Paper coinciding with Unit 3, annotated ……………………………………………………………………….
8
Example Research Paper coinciding with Unit 3 …………………………………………………………………………………….
22
Documenting Research Guide Last Revised: 12/27/2020 2
Outline Structure and Content
The outline is an organization document to provide structure for the research paper. Use the outline to document
research
.
Section 1, Level 1 Section Heading: This heading is the title of the paper.
• Background, topic, introduction
• Describe the broader context in which the problem exists, the topic
• Lead the reader to the problem statement
• Do not explicitly state the problem, research questions, or methodology
This section introduces the research topic and provides a high-level summary of what the reader can
expect to find in the rest of the paper.
Section 1, Level 2 Section Heading:
Statement of the Problem
• This section may come straight from an assignment’s instructions
• Provide the ideal, current, and intent of the problem for research
Section 2, Level 1 Section Heading:
Research Methodology
• Begins with an introduction to all the content in the research methodology section
Section 2, Level 2 Section Heading: Research Questions
• This may come straight from the assignment’s instructions
• Ensure that developed questions conform to the standards defined in the first lecture
Section 2, Level 2 Section Heading:
Sample Data
• Review the sample data – variable names do not identify what the content represents –
so do not use variable names!
• Explain and describe what each of the variables represents, connecting the sample to the
background, problem, and question so the reader can understand what the data
represents and why it is suitable data to answer the research questions
Section 2, Level 2 Section Heading:
Analysis Method and Limitations
• A plan, defining what type of analysis will address each research question.
• The plan will include statistical assumptions, limitations to the analysis method, and
mitigating steps taken for the limitations.
• This section is not a programming plan! This section does not include the programming
procedure or steps. Define this section before conducting any programming or
analysis.
This section finishes with a summary of the content of the section
Develop everything above this statement is before analysis. Work on everything below after the analysis.
EXCEPTION: Develop the reference section before and after analysis.
Some of the elements above this statement could change after analysis.
Section 3, Level 1 Section Heading:
Results
Section 4, Level 1 Section Heading: Discussion
Documenting Research Guide Last Revised: 12/27/2020 3
Section 5, Level 1 Section Heading:
Recommendations for Future Research
Section 6, Level 1 Section Heading:
Conclusion
Section 7, Level 1 Section Heading:
References
• Reference section and references per APA 7
• Some of the standards for this section per APA 7
o References always begin on a new page
▪ use insert new page to ensure this section starts at the top of a separate page from the rest
of the document
o References are in alphabetical order
o Annotated with a hanging indent
▪ The reference begins flush with the left one-inch margin
▪ Indent wrapped text is one-half inch
Documenting Research Guide Last Revised: 12/27/2020 4
The 2016 Presidential Campaign Polling
• The 2016 election was tumultuous
o Distinct perception Trump would not win
o Bias may have played a part
o Polling samples
o shy voters
• The research includes analysis of the polls’ results and how the
results relate to the outcome of the
election.
Statement of the Problem
Neutral polling, collected from a sample genuinely representative of
the voters, will provide an accurate prediction of the winner of an
election. Polling seemed to indicate that Clinton was going to win,
but the electoral vote significantly favored the Trump campaign.
Exploration of the polling results throughout the campaign and a
particularly close look at the ratings at the end of the campaign may
provide insight into the source of the significantly different outcome
than the media portrayed with the election of President Trump.
Research Methodology
Research Question
Considering the 2016 presidential campaign, using the polling data
consolidated by Silver et al. (2016) and the election results
consolidated by Ballotpedia. (n.d.), what relationships exist between
the polling and
the 2016 election results that indicate that President
Trump would win
the election?
Sample Data
Note: Keep in mind that if the data used in an assignment has
variables not used in the analysis, those variables are not part of the
sample! Take note of this in the data. There are several fields not
discussed here, because the fields were not part of the analysis
• The secondary sample data from Silver et al. (2016) includes
polling
data that represents
o Location: fifty states, national polls, and Washington
DC
o Dates: November 2015 to November 2016, the ending
date for each poll
o Size: the sample size of each poll
The title is capitalized in
title case. This is the
first section heading and
the title of the paper in
the final document.
For most of the course this is
provided. In the outline and
research paper, the entire
statement is provided.
Cite the source(s) of the
sample data.
Provide a summary of the
document in the introduction.
While the outline has sentence fragments and bullets throughout – the research paper will not. The
organizational statements in the outline are written as well-developed paragraphs in the research paper.
In APA 7, a level 1
section heading is in
bold, centered between
the one inch left and
right margins.
In APA 7, a level 2
section heading is in
bold, flush to the one
inch left margin.
All research questions
belong in the outline.
Explain the sample in
words.
Explain how the data is
represented, such as parts
per million or percentage
of votes.
Documenting Research Guide Last Revised: 12/27/2020 5
o Vote: the percentage of votes for President Trump
and for Clinton each poll in the data
• The secondary sample data used from Ballotpedia (n.d.)
represents:
o fifty states and Washington DC
o electoral votes available in each state
o 2016 election vote percentage of each state for
President Trump and
Clinton
Analysis Method and Limitations
• What relationships exist between the pre-election polling attributes, the
2016 election, and each state’s allocated electoral votes that indicate that
President Trump would
win the election?
• assessed via visual analysis
o not parametric, therefore no statistical assumptions
o limitations of visual analysis
▪ high dimensionality is challenging to assess
▪ possibility of inadequate assessment leading to
incorrect conclusions
▪ the more comparisons, the higher likelihood of false
discoveries (Zhao et
al., 2017)
o mitigation for inadequate
assessment
▪ explore interesting findings via multiple facets, to
ensure adequate assessment
o mitigation for false discoveries
▪ Attempt to view any key finding from multiple
perspectives, to validate the
finding
Develop everything above this statement for the outline, along with the
reference section.
Develop everything below after the analysis, along with the
reference
section. There may be updates to the other sections.
Type of analysis for each research
question; list each question!
Declare how this method can
address each of the research
questions.
Declare any statistical assumptions
for this method of analysis with a
credible reference.
Provide limitations to the method
of analysis and methods to
mitigate limitation if it impacts
the validity or reliability of the
research.
In other words, if the limitation
can lead to incorrect conclusions,
how will correct conclusions be
determined?
Declare the headings for
the remaining fields
The design for analysis
Documenting Research Guide Last Revised: 12/27/2020 6
Results
Discussion
Recommendations for Future Research
Conclusion
References
Ballotpedia. (n.d.). 2016 election results [dataset]. Retrieved July 18, 2020, from
https://docs.google.com/spreadsheets/d/1zxyOQDjNOJS_UkzerorUCf2OAdcMcIQEwRciKuYBIZ4/pu
bhtml?widget=true&headers=false#gid=658726802
Silver, N., Kanjana, J., & Mehta, D. (2016, November 8). Who will win the presidency? Fivethirtyeight: 20
16
Election Forecast. https://projects.fivethirtyeight.com/2016-
election-forecast/
Zhao, Z., De Stefani, L., Zgraggen, E., Binnig, C., Upfal, E. & Kraska, T. (2017). Controlling false discoveries
during interactive data exploration. In Proceedings of the 2017 ACM international conference on
management of data (pp. 527-540). Association for Computing Machinery.
https://doi.org/10.1145/3035918.306401
9
Include the reference(s) of the data, in APA
7.
Include a citation for every
reference
Include a reference for every
citation
The reference section begins on a separate page.
Documenting Research Guide Last Revised: 12/27/2020 7
Writing Tips
• When writing a paper or developing a presentation, always include a summary of the document within
the introduction and the conclusion.
• Focus the writing on the purpose: solve the problem, answer the question, or prove the expected
outcome. In this course, the assignments will all have research questions. Focus on the questions.
• Write concisely. This is not a persuasive paper. Writing superfluously devalues your work.
• When you finish writing:
o Read the document aloud.
▪ This is the single, most effective method to identify elements of the document that
require editing.
▪ Think about the problem, research questions or the expected outcome:
• Did you focus on it throughout the document?
• Did you provide answers to the research question(s)?
o If you are not particularly confident in your writing:
▪ Take time to identify the topic sentence in every paragraph, in every section, and within
the introduction and conclusion.
▪ There should be transition sentences between the ideas in the document. Does the writing
jump from one idea to the next?
▪ The writing center is an excellent resource, as well.
▪ Use the outline to organize your graduate-level writing.
• Do not concern yourself with your SafeAssign score.
o Ensure that quoted words, paraphrasing, and direct references to external sources have citations
and references to the original source of the information. Still not sure? Email me.
o Think about it! What do you think the average SafeAssign percentage is for the outline?
▪ A significant portion of the outline will come from the assignment instructions.
▪ The matching criteria from SafeAssign typically allocates 60-80% scores to submissions
that are correctly written.
• Cite every reference. Include all references in the reference section.
• Evaluation of all writing assignments by APA 7 criteria.
o Student papers do not include an abstract.
o Vertical spacing is uniform between lines of text
▪ Microsoft Word automatically adds paragraph padding – remove it or use the template.
o The text alignment throughout the document is left-align, not justify.
o Do not solely rely on citation and reference generators. These tools are fallible.
8
Documenting Research Guide Last Revised: 12/27/2020 8
The 2016 Presidential Campaign Polling
Dr. Kathy A. McClure
University of the Cumberlands
ITS-530: Data Analysis and Visualization
Dr. Kathy A. McClure
July 23, 20
20
One of two places in the
document correctly
documented with non-
uniform vertical spacing.
.
The top name is author.
When you see my name
again it is for the
professor of the course.
The only element in the header is the page number in the
same font as the document, starting at 1. (As this is part
of an example document the numbering is different.)
There is no footer in the student research paper.
There is no footer in a student research paper, per APA 7.
This footer is for document control of the Documenting Research Guide.
9
Documenting Research Guide Last Revised: 12/27/2020 9
The 2016 Presidential Campaign Polling
The 2016 presidential campaign was tumultuous. It had seemed impossible that President
Trump would win the election. Silver et al. (2016)
indicated that there was a 71.4% chance that Clinton
would win the election. During the campaign, the media
led voters, including elected members of the republican
party, to believe that President Trump would not win the
election (Hohman, 2016). Regardless of the media,
Hohman (2016) retroactively identified that there were
many voters that were not pro-Clinton leading up to the
election. Stevenson (2016) interviewed American
University professor Dr. Allan Lichtman, who overtly
stated that President Trump would win the election based
on historical voting in this country. Dr. Lichtman
specified to exceptions to this claim: candidate Johnson
must receive at least five percent of the vote and
President Trump’s unpredictable behavior. Goldmacher and
Schreckinger (2016) stated that President Trump winning
the election was the “…biggest upset in U.S. history”
(title). Many believed Clinton would win.
Problem Statement
Polling samples that represent the population will
provide an accurate prediction of the election winner.
Note that the outline was not followed
explicitly for the topic/introduction
Don’t forget to cite and reference sources
of information
Use evidence to support any assertions
that are not common knowledge
Example: “Sampling bias was an issue in
all polls.” That statement infers this is a
fact – when it is not and it would be
impossible to prove this statement!
You must have a citation and reference for
assertions.
From the outline:
• The 2016 election was tumultuous
• Distinct perception Trump would
not win
• Bias may have played a part
• Polling samples
• shy voters
• The research includes analysis of
the polls’ results and how the results
relate to the outcome of the election
Why did this quote end with the word “title”
in parentheses? It is cited correctly. The
statement began with the source authors and
date. A quote requires three parts in the
citations, author, data, and the page number.
The reference is a website, so there are no
page or paragraph numbers. It must identify
where the quote was found, in this case, the
title.
The problem statement is verbatim from the
outline, unless it was insufficient.
10
Documenting Research Guide Last Revised: 12/27/2020 10
Polling results appeared to indicate that Clinton was going to win, but the election resulted in
President Trump swearing-in as the 45th president. Exploration of the polling and election results
may provide insight as to why the election winner was unexpected.
Method
Research Question
Considering the 2016 presidential campaign,
using the polling data consolidated by Silver et al. (2016)
and the election results consolidated by Ballotpedia. (n.d.), what relationships exist between the
polling and the 2016 election results that indicate that President Trump would win the election?
Sample
This research employed two secondary data sources
for the analysis. Consolidated polling data collected by
Silver et al. (2016) is the first data source. Each observed
poll includes the percentage of votes by location, ending
date, and sample size for Clinton and President Trump.
Ballotpedia (n.d.) election data is also necessary for this
analysis and includes the percentage of votes by location
for Clinton and President Trump. Available electoral
votes for each location is another attribute in the election
data. Locations between the two secondary data sources
differed.
The polls’ locations include the entire nation, each
state, and Washington, DC, and specific districts within Nebraska and Maine. The district polls
The research question(s) are verbatim from
the outline unless the question was
insufficient.
From the outline:
The secondary sample data from
Silver et al. (2016) includes polling
data that represents
• fifty states, national polls, and
Washington DC
• November 2015 to November
2016, the ending date for each poll
• the sample size of each poll
• provides a raw percentage of votes
for each poll for President Trump
and Clinton
The secondary sample data used from
Ballotpedia (n.d.) represents:
• fifty states and Washington DC
• electoral votes available in each
state
• 2016 election vote percentage of
each state for President Trump and
Clinton
11
Documenting Research Guide Last Revised: 12/27/2020 11
within Nebraska and Maine were representative of the method of electoral vote distribution.
Splitting the electoral vote is possible in Nebraska and Maine (Coleman, 2020). In the other 48
states and Washington, DC, using winner-take-all, the
popular vote winner for the state receives all the electoral
votes. The election data simplified the locations: each
state and Washington, DC.
Analysis Method and Limitations
The method of analysis must be suitably capable
of meeting the objective of this research, statistical
assumptions identification is necessary, if they exist, and
identification of any limitations is essential, along with
mitigation, where possible. Visual analysis is suitable for
extracting relationships that may exist in the data. This
method is also appropriate for confirming the information
derived from the analysis. There are no formal statistical
assumptions. There are three limitations identified for visual
analysis.
High dimensionality, inadequate assessment, and false discoveries are risks associated
with visual analysis. The scope of this research does not include numerous variables, mitigating
the threats associated with high dimensionality. The potential for inadequate assessment and
false discoveries requires mitigation. Visualizations of data provide a perspective of the
information without context. To mitigate these risks, it is compulsory to assess all key findings
from multiple perspectives. This process ensured that there was an adequate assessment of that
From the outline:
Analysis Method and Limitations
• assessed via visual analysis
• not parametric, therefore no statistical
assumptions
• limitations of visual analysis
• high dimensionality is challenging to
assess
• possibility of inadequate assessment
leading to incorrect conclusions
• the more comparisons, the higher
likelihood of false discoveries (Zhao et
al., 2017)
• mitigation for inadequate assessment
• explore interesting findings via
multiple facets, to ensure adequate
assessment
• mitigation for false discoveries
• Attempt to view any key finding from
multiple perspectives, to validate the
finding
12
Documenting Research Guide Last Revised: 12/27/2020 12
the perceived information. Focusing on the research question and using two sources of secondary
data, the analysis generated results.
Results
Consolidation of the visual analysis highlighted key findings through four visualizations
of data. Manipulating the data with various summarization techniques generated meaningful
graphics. The sample included nearly a year’s worth of polling data, but limiting the data to polls
closest to the election generated the key findings in this research. The term polling vote
represents polls ending in November 2016, consolidated by state and candidate, using the median
value. Geospatial visualization indicates that in 45 of the 50 states the winning candidate in the
polling vote and the election were the same (see Figure 1). In five states, Clinton led in the
polling vote, but President Trump won in the
election. For simplification, the term flipped states
refers to the five states identified in Figure 1.
Due to the non-uniformity of the data, the measure of centrality in this analysis is the
median. Summarizing data can cause misrepresentation of the data. Comparing the polling vote
identified 12 states with five percent or less difference between candidates. Visualizing the 12
states identified the how well the median represents the data (see Figure 2). The evidence
suggests that the median does not misrepresent the results. The 12 states include the five flipped
states identified in Figure 1. The close margins in the polling data of the flipped states
necessitated a deeper investigation, into individual polls. Before documenting the remaining
results of this analysis, the visualization of the difference between candidates requires further
explanation.
Repeating the same information is ill-advised.
Don’t repeat the information in the caption of a
figure or table.
13
Documenting Research Guide Last Revised: 12/27/2020 13
The candidates were compared by subtracting the polling votes for each state (see Figure
3 and Figure 4). The values’ direction is indicative of the winning candidate. Leads held by
Clinton are to the left of zero. Where President Trump’s led, the value is annotated to the right of
zero. The value is indicative of how much lead one candidate has over the other. For example, if
President Trump earned 40% of the vote and Clinton earned 41% of the vote, Clinton led that
vote by one percent. This Clinton lead would be visualized by placing the marker to the left of
zero on the axis marker representing a value of one percent.
APA use of figures & tables is specific. Each figure or table but include enough information to be self-explanatory.
Do not explain the figure in the document. **You must refer to each figure or table in the document, though!**
Results require EVIDENCE. In visual analysis, the evidence is visual!
14
Documenting Research Guide Last Revised: 12/27/2020 14
15
Documenting Research Guide Last Revised: 12/27/2020 15
16
Documenting Research Guide Last Revised: 12/27/2020 16
17
Documenting Research Guide Last Revised: 12/27/2020 17
After identifying the flipped states’ polling vote by candidate differed by five percent or
less, each poll within flipped states ending in November 2016 were analyzed (see Figure 3). The
majority of the individual polls also varied by less than five percent
between the candidates. Clinton held the lead in nearly all polls in
these states. In Florida, there were no polls that exceeded the five
percent margin between candidates. Trump did not lead in any polls
in
Wisconsin from this data.
The polling vote and election vote were compared by
candidate all election locations from the data. While five states
flipped, there were other states with close margins. Additionally, the
comparison of the polling vote and election vote visualizes the relationship between the
candidates’ polling vote and the election vote (see Figure 4). The 12 states shown in Figure 2, are
annotated with green text in Figure 4.
Discussion
Q1. Considering the 2016 presidential campaign,
using the polling data consolidated by Silver et al. (2016)
and the election results consolidated by Ballotpedia.
(n.d.), what relationships exist between the polling and
the 2016 election results that indicate that President
Trump would win the election?
The close margins in multiple states in the polling
data indicate that the candidates between candidates
suggest that there were no guarantees in this election.
What is the difference between
an assessment and an assertion?
“I am short” – assessment
“I am 5’6” – assertion
Which one requires evidence?
Every assertion, that is not
common knowledge.
What evidence?
Evidence is derived from
the analysis or
a cited reference.
How did this example begin?
The research question!
Okay, how did this example begin after
the research question?
The close margins in multiple states…
That is the topic sentence for the section.
What can you expect to find in this
section?
Did you notice that this section isn’t all
that long? There are not a lot of findings to
discuss in regards to the research question.
18
Documenting Research Guide Last Revised: 12/27/2020 18
Florida voting was amongst the closest margins in both the polls and in the election (see Figure
4). As a state, Florida has 29 winner-take-all electoral votes and the polling margins were small
enough to state that any uncertainty would indicate that the polling results were not able to
identify a winner. While 29 votes would not have changed the outcome, this state was not the
only state with close margins. Amongst the polling votes, 12 states, representing over 100
electoral votes, had margins of less than five percent between President Trump and Clinton. It is
reasonable to assume that polls are not perfect. The possibility of underrepresenting a genre of
the population is too great of a possibility. Through visual analyses, this evidence suggests that
either candidate could have
won the election due to uncertainty.
Recommendations for Future Research
Two identified opportunities may provide more insight into why President Trump won
the election, despite the low likelihood identified by analysts such as Silver et al. (2016). The
polling vote for President Trump is underrepresented in many of the states where he held the lead
(see Figure 4). Conversely, in states that Clinton led the polling vote represents the election
reasonably well. Kurtzleben (2016) did some analysis in this area and inferred that rural voting
was pro-Trump. Analysis conducted by Lee (2017) investigated the impact of rural and urban
voters in the 2016 election. Lee’s analysis of voting data in Minnesota and Wisconsin suggests
that urban area voters were strong supporters of Clinton, and rural voters were strong supporters
of President Trump. The dispersion of rural and urban voters may not be recoverable for this
polling data. Uncovering the source of the underrepresented President Trump vote could indicate
a systemic issue in polling conducted in the 2016 presidential election. With additional data, the
first recommendation for future research is to identify poll and election votes that were allocated
to either rural or urban votes. The confidence interval is a statistical measure of uncertainty.
19
Documenting Research Guide Last Revised: 12/27/2020 19
Reassessing this data, implementing poll confidence intervals into an analysis method capable of
prediction is the second recommendation for future research. Either of these research
opportunities could add more insight into the disparity between polling and the 2016 presidential
election.
Conclusion
Assessing relationships in the polling data and election data for the 2016 presidential
election, indicates that due to uncertainty the winner of the
election could not be reasonably determined. Uncertainty in the
polling data and close margins between candidates suggest neither
candidate held the lead. Electoral votes allocated to states with
close margins, along with the even split of states between Clinton and President Trump, suggests
that there is insufficient evidence to determine the likelihood of an election winner. Perhaps
analysts should have followed the method used by Dr. Lichtman when he stated that President
Trump would win
(Stevenson, 2016). Afterall, Dr. Lichtman was correct.
This section should summarize the
entire document.
This section may highlight key
findings, as well.
No new information!
20
Documenting Research Guide Last Revised: 12/27/2020 20
References
Ballotpedia. (n.d.). 2016 election results [dataset]. Retrieved July 18, 2020, from
https://docs.google.com/spreadsheets/d/1zxyOQDjNOJS_UkzerorUCf2OAdcMcIQEwRc
iKuYBIZ4/pubhtml?widget=true&headers=false#gid=658726802
Coleman, J. M. (2020, January 9). The electoral college: Maine and Nebraska’s crucial
battleground votes. Sabato’s Crystal Ball.
http://centerforpolitics.org/crystalball/articles/the-electoral-college-maine-and-nebraskas-
crucial-battleground-votes/
Goldmacher, S., & Schreckinger, B. (2016, November 16). Trump pulls off biggest upset in U.S.
history. Politico. https://www.politico.com/story/2016/11/election-results-2016-clinton-
trump-231070
Hohman, J. (2016, November 9). The daily 202: Why Trump won — and why the media missed it.
The Washington Post. https://www.washingtonpost.com/news/powerpost/paloma/daily-
202/2016/11/09/daily-202-why-trump-won-and-why-the-media-missed-
it/5822ea17e9b69b6085905dee/
Kurtzleben, D. (2016, November 14). Rural voters played a big part in helping Trump defeat
Clinton. NPR. https://www.npr.org/2016/11/14/501737150/rural-voters-played-a-big-
part-in-helping-trump-defeat-clinton
Lee, M. (2017, January 5). Mapping Wisconsin presidential election results [web log]. Retrieved
August 21, 2020, from https://www.mikelee.co/posts/2016-12-26-wisconsin-presidential-
election-results/
21
Documenting Research Guide Last Revised: 12/27/2020 21
Silver, N., Kanjana, J., & Mehta, D. (2016, November 8). Who will win the presidency?
Fivethirtyeight: 2016 Election Forecast. https://projects.fivethirtyeight.com/2016-
election-forecast/
Stevenson, P. W. (2016, November 9). Professor who predicted 30 years of presidential
elections correctly called a Trump win in September. The Washington Post.
https://www.washingtonpost.com/news/the-fix/wp/2016/10/28/professor-whos-predicted-
30-years-of-presidential-elections-correctly-is-doubling-down-on-a-trump-win/
Zhao, Z., De Stefani, L., Zgraggen, E., Binnig, C., Upfal, E. & Kraska, T. (2017). Controlling
false discoveries during interactive data exploration. In Proceedings of the 2017 ACM
international conference on management of data (pp. 527-540). Association for
Computing Machinery. https://doi.org/10.1145/3035918.3064019
Pay attention to the formatting here! APA 7 is not the same as APA 6.
Every reference MUST be in a citation somewhere in the text document.
When do you cite? Paraphrasing, quoting, or direct reference to a source.
When annotating references:
• Every reference has an author – the author may not be a person
• Every reference has a date – more often than not, it is only the year
• Every reference has a title – unless the title is the author!
• Every reference includes a source.
o Webpages are sourced from websites.
o Journal articles are sourced from journals.
o PUBLISHED conference papers are from proceedings from a publisher (so
both are needed! – see the reference to Zhao et al.)
o Conference papers are from conferences, when they are not published.
• Unless there is no electronic version – every reference has a “home” that is included.
o IF a DOI exists it must be the link via the DOI.
o The website is not optional.
22
Documenting Research Guide Last Revised: 12/27/2020 22
23
Documenting Research Guide Last Revised: 12/27/2020 23
The 2016 Presidential Campaign Polling
Dr. Kathy A. McClure
University of the Cumberlands
ITS-530: Data Analysis and Visualization
Dr. Kathy A. McClure
July 23, 2020
24
Documenting Research Guide Last Revised: 12/27/2020 24
The 2016 Presidential Campaign Polling
The 2016 presidential campaign was tumultuous. It had seemed impossible that President
Trump would win the election. Silver et al. (2016) indicated that there was a 71.4% chance that
Clinton would win the election. During the campaign, the media led voters, including elected
members of the republican party, to believe that President Trump would not win the election
(Hohman, 2016). Regardless of the media, Hohman (2016) retroactively identified that there
were many voters that were not pro-Clinton leading up to the election. Stevenson (2016)
interviewed American University professor Dr. Allan Lichtman, who overtly stated that
President Trump would win the election based on historical voting in this country. Dr. Lichtman
specified to exceptions to this claim: candidate Johnson must receive at least five percent of the
vote and President Trump’s unpredictable behavior. Goldmacher and Schreckinger (2016) stated
that President Trump winning the election was the “…biggest upset in U.S. history” (title). Many
believed that Clinton would win.
Problem Statement
Polling samples that represent the population will provide an accurate prediction of the
election winner. Polling results appeared to indicate that Clinton was going to win, but the
election resulted in President Trump swearing-in as the 45th president. Exploration of the polling
and election results may provide insight as to why the election winner was unexpected.
Method
Research Question
Considering the 2016 presidential campaign, using the polling data consolidated by Silver
et al. (2016) and the election results consolidated by Ballotpedia. (n.d.), what relationships exist
25
Documenting Research Guide Last Revised: 12/27/2020 25
between the polling and the 2016 election results that indicate that President Trump would win
the election?
Sample
This research employed two secondary data sources for the analysis. Consolidated polling
data collected by Silver et al. (2016) is the first data source. Each observed poll includes the
percentage of votes by location, ending date, and sample size for Clinton and President Trump.
Ballotpedia (n.d.) election data is also necessary for this analysis and includes the percentage of
votes by location for Clinton and President Trump. Available electoral votes for each location is
another attribute in the election data. Locations between the two secondary data sources differed.
The polls’ locations include the entire nation, each state, and Washington, DC, and
specific districts within Nebraska and Maine. The district polls within Nebraska and Maine were
representative of the method of electoral vote distribution. Splitting the electoral vote is possible
in Nebraska and Maine (Coleman, 2020). In the other 48 states and Washington, DC, using
winner-take-all, the popular vote winner for the state receives all the electoral votes. The election
data simplified the locations: each state and Washington, DC.
Analysis Method and Limitations
The method of analysis must be suitably capable of meeting the objective of this
research, statistical assumptions identification is necessary, if they exist, and identification of any
limitations is essential, along with mitigation, where possible. Visual analysis is suitable for
extracting relationships that may exist in the data. This method is also appropriate for confirming
the information derived from the analysis. There are no formal statistical assumptions. There are
three limitations identified for visual analysis.
26
Documenting Research Guide Last Revised: 12/27/2020 26
High dimensionality, inadequate assessment, and false discoveries are risks associated
with visual analysis. The scope of this research does not include numerous variables, mitigating
the threats associated with high dimensionality. The potential for inadequate assessment and
false discoveries requires mitigation. Visualizations of data provide a perspective of the
information without context. To mitigate these risks, it is compulsory to assess all key findings
from multiple perspectives. This process ensured that there was an adequate assessment of that
the perceived information. Focusing on the research question and using two sources of secondary
data, the analysis generated results.
Results
Consolidation of the visual analysis highlighted key findings through four visualizations
of data. Manipulating the data with various summarization techniques generated meaningful
graphics. The sample included nearly a year’s worth of polling data, but limiting the data to polls
closest to the election generated the key findings in this research. The term polling vote
represents polls ending in November 2016, consolidated by state and candidate, using the median
value. Geospatial visualization indicates that in 45 of the 50 states the winning candidate in the
polling vote and the election were the same (see Figure 1). In five states, Clinton led in the
polling vote, but President Trump won in the election. For simplification, the term flipped states
refers to the five states identified in Figure 1.
Due to the non-uniformity of the data, the measure of centrality in this analysis is the
median. Summarizing data can cause misrepresentation of the data. Comparing the polling vote
identified 12 states with five percent or less difference between candidates. Visualizing the 12
states identified the how well the median represents the data (see Figure 2). The evidence
suggests that the median does not misrepresent the results. The 12 states include the five flipped
27
Documenting Research Guide Last Revised: 12/27/2020 27
states identified in Figure 1. The close margins in the polling data of the flipped states
necessitated a deeper investigation, into individual polls. Before documenting the remaining
results of this analysis, the visualization of the difference between candidates requires further
explanation.
28
Documenting Research Guide Last Revised: 12/27/2020 28
29
Documenting Research Guide Last Revised: 12/27/2020 29
30
Documenting Research Guide Last Revised: 12/27/2020 30
31
Documenting Research Guide Last Revised: 12/27/2020 31
The candidates were compared by subtracting the polling votes for each state (see Figure
3 and Figure 4). The values’ direction is indicative of the winning candidate. Leads held by
Clinton are to the left of zero. Where President Trump’s led, the value is annotated to the right of
zero. The value is indicative of how much lead one candidate has over the other. For example, if
President Trump earned 40% of the vote and Clinton earned 41% of the vote, Clinton led that
vote by one percent. This Clinton lead would be visualized by placing the marker to the left of
zero on the axis marker representing a value of one percent.
After identifying the flipped states’ polling vote by candidate differed by five percent or
less, each poll within flipped states ending in November 2016 were analyzed (see Figure 3). The
majority of the individual polls also varied by less than five percent between the candidates.
Clinton held the lead in nearly all polls in these states. In Florida, there were no polls that
exceeded the five percent margin between candidates. Trump did not lead in any polls in
Wisconsin from this data.
The polling vote and election vote were compared by candidate all election locations
from the data. While five states flipped, there were other states with close margins. Additionally,
the comparison of the polling vote and election vote visualizes the relationship between the
candidates’ polling vote and the election vote (see Figure 4). The 12 states shown in Figure 2, are
annotated with green text in Figure 4.
Discussion
Q1. Considering the 2016 presidential campaign, using the polling data consolidated by
Silver et al. (2016) and the election results consolidated by Ballotpedia. (n.d.), what relationships
exist between the polling and the 2016 election results that indicate that President Trump would
win the election?
32
Documenting Research Guide Last Revised: 12/27/2020 32
The close margins in multiple states in the polling data indicate that the candidates
between candidates suggest that there were no guarantees in this election. Florida voting was
amongst the closest margins in both the polls and in the election (see Figure 4). As a state,
Florida has 29 winner-take-all electoral votes and the polling margins were small enough to state
that any uncertainty would indicate that the polling results were not able to identify a winner.
While 29 votes would not have changed the outcome, this state was not the only state with close
margins. Amongst the polling votes, 12 states, representing over 100 electoral votes, had margins
of less than five percent between President Trump and Clinton. It is reasonable to assume that
polls are not perfect. The possibility of underrepresenting a genre of the population is too great
of a possibility. Through visual analyses, this evidence suggests that either candidate could have
won the election due to uncertainty.
Recommendations for Future Research
Two identified opportunities may provide more insight into why President Trump won
the election, despite the low likelihood identified by analysts such as Silver et al. (2016). The
polling vote for President Trump is underrepresented in many of the states where he held the lead
(see Figure 4). Conversely, in states that Clinton led the polling vote represents the election
reasonably well. Kurtzleben (2016) did some analysis in this area and inferred that rural voting
was pro-Trump. Analysis conducted by Lee (2017) investigated the impact of rural and urban
voters in the 2016 election. Lee’s analysis of voting data in Minnesota and Wisconsin suggests
that urban area voters were strong supporters of Clinton, and rural voters were strong supporters
of President Trump. The dispersion of rural and urban voters may not be recoverable for this
polling data. Uncovering the source of the underrepresented President Trump vote could indicate
a systemic issue in polling conducted in the 2016 presidential election. With additional data, the
33
Documenting Research Guide Last Revised: 12/27/2020 33
first recommendation for future research is to identify poll and election votes that were allocated
to either rural or urban votes. The confidence interval is a statistical measure of uncertainty.
Reassessing this data, implementing poll confidence intervals into an analysis method capable of
prediction is the second recommendation for future research. Either of these research
opportunities could add more insight into the disparity between polling and the 2016 presidential
election.
Conclusion
Assessing relationships in the polling data and election data for the 2016 presidential
election, indicates that due to uncertainty the winner of the election could not be reasonably
determined. Uncertainty in the polling data and close margins between candidates suggest
neither candidate held the lead. Electoral votes allocated to states with close margins, along with
the even split of states between Clinton and President Trump, suggests that there is insufficient
evidence to determine the likelihood of an election winner. Perhaps analysts should have
followed the method used by Dr. Lichtman when he stated that President Trump would win
(Stevenson, 2016). Afterall, Dr. Lichtman was correct.
34
Documenting Research Guide Last Revised: 12/27/2020 34
References
Ballotpedia. (n.d.). 2016 election results [dataset]. Retrieved July 18, 2020, from
https://docs.google.com/spreadsheets/d/1zxyOQDjNOJS_UkzerorUCf2OAdcMcIQEwRc
iKuYBIZ4/pubhtml?widget=true&headers=false#gid=658726802
Coleman, J. M. (2020, January 9). The electoral college: Maine and Nebraska’s crucial
battleground votes. Sabato’s Crystal Ball.
http://centerforpolitics.org/crystalball/articles/the-electoral-college-maine-and-nebraskas-
crucial-battleground-votes/
Goldmacher, S., & Schreckinger, B. (2016, November 16). Trump pulls off biggest upset in U.S.
history. Politico. https://www.politico.com/story/2016/11/election-results-2016-clinton-
trump-231070
Hohman, J. (2016, November 9). The daily 202: Why Trump won — and why the media missed it.
The Washington Post. https://www.washingtonpost.com/news/powerpost/paloma/daily-
202/2016/11/09/daily-202-why-trump-won-and-why-the-media-missed-
it/5822ea17e9b69b6085905dee/
Kurtzleben, D. (2016, November 14). Rural voters played a big part in helping Trump defeat
Clinton. NPR. https://www.npr.org/2016/11/14/501737150/rural-voters-played-a-big-
part-in-helping-trump-defeat-clinton
Lee, M. (2017, January 5). Mapping Wisconsin presidential election results [web log]. Retrieved
August 21, 2020, from https://www.mikelee.co/posts/2016-12-26-wisconsin-presidential-
election-results/
35
Documenting Research Guide Last Revised: 12/27/2020 35
Silver, N., Kanjana, J., & Mehta, D. (2016, November 8). Who will win the presidency?
Fivethirtyeight: 2016 Election Forecast. https://projects.fivethirtyeight.com/2016-
election-forecast/
Stevenson, P. W. (2016, November 9). Professor who predicted 30 years of presidential
elections correctly called a Trump win in September. The Washington Post.
https://www.washingtonpost.com/news/the-fix/wp/2016/10/28/professor-whos-predicted-
30-years-of-presidential-elections-correctly-is-doubling-down-on-a-trump-win/
Zhao, Z., De Stefani, L., Zgraggen, E., Binnig, C., Upfal, E. & Kraska, T. (2017). Controlling
false discoveries during interactive data exploration. In Proceedings of the 2017 ACM
international conference on management of data (pp. 527-540). Association for
Computing Machinery. https://doi.org/10.1145/3035918.3064019
- Documenting Research Guide
Outline Structure and Content
Outline Example: based on the analysis in Unit 3
Writing Tips
1/13/21Assignment 1 ID NH ND x P a g e | 1
Research Assignment 1
The first week you will submit an outline based on the instructions. The following week you will do
this assignment, submitting a paper and an R script file. Look at the examples in the Documenting
Research Guide before reading through these instructions. Ask questions, if needed!
The data consolidated by the Centers for Disease Control and Prevention (CDC) is used to
determine the most vulnerable areas should a disaster occur. In a perfect world, vulnerability
indicators would represent the people correctly. Currently, this far-from-perfect method is the best
that has been developed. There may be indicators that are not adequately predictive of social
vulnerability. Understanding the influence of these attributes can improve the assessment,
improving the ability to predict the impact of disasters on individual communities.
What relationships exist in the states of Idaho, New Hampshire, and North Dakota between
the
socioeconomic fields, household composition and disability fields, and the estimated number of
minorities, the estimated number of homes with no vehicle, and the tract population, and the social
vulnerability index when using the data consolidated by the CDC (n.d.)?
What indicators in the states of Idaho, New Hampshire, and North Dakota between the
socioeconomic fields, household composition and disability fields, and the estimated number of
minorities, the estimated number of homes with no vehicle, and tract population have the most
influence in predicting social vulnerability when using the data consolidated by the CDC (n.d.)?
• The secondary data reference is below, formatted per APA 7. Update the retrieval date to the date
you retrieved it:
Centers for Disease Control and Prevention. (n.d.). CDC social vulnerability index 2018 US [Data set and
code book]. Agency for Toxic Substances and Disease Registry. Geospatial Research,
, and
Services Program. Retrieved January 4, 2021, from
https://www.atsdr.cdc.gov/placeandhealth/svi/data_documentation_download.html
• The data directly:
https://svi.cdc.gov/Documents/Data/2018_SVI_Data/CSV/SVI2018_US.csv
• The data dictionary or code book directly:
https://svi.cdc.gov/Documents/Data/2018_SVI_Data/SVI2018Documentation
• Create a subset of the data to represent the secondary data sample for this analysis.
• Don’t include observations with a total population of zero in your analysis. Think about it; if there’s
no population, how can risk to the community be assigned?
• There are 13 variables used in this analysis. When you write about the secondary data sample,
you only need to discuss the data you used. If observations (rows of data) were excluded, that
needs to be discussed. Cite and reference sources that you use to identify variable content.
• Do not use more than one field for each variable. Other than the field that represents the SVI, all
of your variables are prefixed with E_. For example, there are multiple fields with “PCI” for per
capita income, but only one E_PCI.
• Don’t copy and paste the following data sample information into your outline. It’s insufficient.
• How do you know what data to use? It’s in the research question.
Do not
modify the
data outside
of R.
1/13/21 Assignment 1 ID NH ND x P a g e | 2
o socioeconomic fields
▪ estimated
quantities of:
o people living below the poverty level
o people unemployed
o people without a high school diploma
▪ tract average per capita income
o household composition and disability fields, also estimated
quantities of:
▪ people age 65 and over
▪ people age 17 and under
▪ disabled
▪ single-parent homes with children under 18
o estimated number of minorities
o estimated number of homes with no vehicle
o estimated tract population
o the SVI index is RPL_THEMES, in column 99
o the state
• It is unlikely that any action taken in cleaning is documented in your research paper. If these steps
were documented in a paper, they would be a part of the procedures section. I don’t require you to
write the procedures section because you submit an R file.
• When changing an object or part of an object, validate every change, and comment in your code.
• There is a code representing missing values; use the data dictionary to learn more! Reassign the
values as NA, if any observations in your sample data include this code.
Analysis
• Conduct two types of analysis: visual analysis to identify relationships and a random forest model
to identify the indicators’ influence in predicting the SVI.
• Connect the relationships and influence measures aforementioned to the research questions when
you document your Analysis Methods and Limitations section.
Results section and the discussion section
• During the visual analysis, only present meaningful visuals in your paper. Provide your
interpretations of any results you present.
• Ensure you establish that the model is valid and reliable in your documentation before discussing
the influence the different fields have on predicting the outcome.
Using the first research question, the variables are in red:
What relationships exist in the states of
New Hampshire, North Dakota, and South Dakota
between the
socioeconomic fields,
household composition and disability fields, and
the estimated number of minorities,
the estimated number of homes with no vehicle,
and the
tract population, and the
social vulnerability index
when using the data consolidated by the CDC (n.d.)?
Use the data dictionary to uncover which variables in the data align with these
variables. Look at the example information from the data dictionary in the two
partial images to the right.
Modified from CDC (n.d., p. 5)
Modified from CDC (n.d., p. 6)
1/13/21 Assignment 1 ID NH ND x P a g e | 3
• Do not speculate. Use evidence. When documenting the results, consider the generalizability.
• Your interpretations of your results are crucial to demonstrating your understanding.
Future recommendations section
• Include recommendations for future analysis, based on your research in R.
• An example future research recommendation may look something like this:
An opportunity for future research is exploration modeling to determine what other
variables, when eliminated, have little or no impact when predicting the SVI based on the
supporting characteristics in the data.
Create a random forest model for each state that is assigned. You will need to write a research question
that aligns with the problem statement, providing your objective of these state-level models. What is it
that you are looking for? The objective can be the same as the second research question in these
instructions or one you develop independently. Use the criteria found in Unit 1 Part 1 to make sure your
research question is sound. Want to try the challenge, but need help? Please email me.
• The week you initially receive these instructions, the objective is to complete an outline. Use these
instructions, the data, the data dictionary, and the Documenting Research Guide to complete the
outline.
o Submit as an MS Word document file type
▪ The formatting is not crucial.
▪ HINT Most of the outline is copied from the instructions. Focus on what you write.
▪ Don’t forget to cite and reference any sources you use to complete the outline.
• The second week you receive these instructions, you will complete this assignment and submit:
o Submit as an MS Word document file type
▪ Adhere to the standards of APA 7
▪ Use the Student Paper Template in the Useful Documents folder in Blackboard; it’s
preformatted per APA 7.
▪ Length 3-5 pages and at least 1000 words in the body of the document; count
excludes the cover page, tables, or figures, or the reference page.
o R Script; the final version in a .R file type
• See the Documenting Research Guide for more details on what is required.
• Questions? Please email me. Stuck on the programming or paper? Please email me.
• You will receive an error notification when you submit because of the .R file type. Check your
email for the submission confirmation email automatically sent from Blackboard.
• Ensure that every reference in the reference list is also cited in the text.
• Do not forget to cite and reference the source of the data.
• Use the problem statement and research questions verbatim as in these instructions.
• If your submission adheres to a version of this assignment not available to you in Blackboard, you
will earn a zero and be documented as demonstrating academic dishonesty.
• This is an individual assignment. Do not share your work and don’t accept others’ work.
• Take a look at the rubric to get the best possible grade.
1/13/21 Assignment 1 ID NH ND x P a g e | 4
References
Centers for Disease Control and Prevention. (n.d.). CDC social vulnerability index 2018 US
[Data set and code book]. Agency for Toxic Substances and Disease Registry. Geospatial
Research, Analysis, and Services Program. Retrieved January 4, 2021, from
https://www.atsdr.cdc.gov/placeandhealth/svi/data_documentation_download.html
Flanagan, B. E., Gregory, E. W., Hallisey, E. J., Heitgerd, J. L., and Lewis, B. (2011). A social
vulnerability index for disaster management. Journal of Homeland Security and
Emergency Management, 8(1), 1-22. https://doi.org/10.2202/1547-7355.1792
- Problem
Question 1
Question 2
Data
Collecting data
Data cleaning
Analysis
When writing your paper
Results section and the discussion section
Future recommendations section
Extra credit challenge
Required files to submit for this assignment
Important Information