Project 2 Outline

Part 1: Data Source

• I’m using a couple different data sets at the moment to create a combined data set for Pennsylvania and its respective counties. The URL’s are listed below for where I downloaded my files. With these sets I want to look at current polling and prior election data for each county and if factors like education, poverty, unemployment, and more have an effect or correlation on polling or voting trends. I am at the moment having trouble finding polling by county for PA. I had to scrap my previous idea for the project and adjust accordingly. This is an extremely rough outline as to my plan and I understand I need much more.

Census Site | https://hdpulse.nimhd.nih.gov/data-portal/social/map?socialtopic=090&socialtopic_options=social_6&demo=00012&demo_options=workforce_2&race=00&race_options=raceall_1&sex=0&sex_options=sex_3&age=916&age_options=age16_1&statefips=42&statefips_options=area_states

Election Results 2020 | https://www.electionreturns.pa.gov/General/SummaryResults?ElectionID=83&ElectionType=G&IsActive=0

• Overall the data is decent and clean apart from a small couple of small fixes needed data. For the election data there is missing info but this do to the recording data that was not relevant in this case. They recorded Yes or No votes which you wouldn’t need when choosing a particular candidate.

From the 2020 Election Results:

• County Name: This is just a designation for each county

• Candidate Name: The candidate running

• Votes: Total Votes (By each candidate)

• Election Day Votes: Votes on Election day

• Mail Votes: Votes by mail in ballot

• Provisional Votes: Provisional ballot votes for those with unsure elegibility

From the Census Data Overall:

• County: Name of County

• FIPS: Number Identifier for County

From Bachelors Data:

• Value (percent): Percent of people in county with degree

• People: People with bachelors by county

From Unemployed Data:

• Value (percent): Percent of people unemployed per county

• People: People unemployed by county

From Poverty Data:

• Value (percent): Percent of poverty per county

• Families: Families in poverty by county

Once I find a solid polling data by county, I will have more columns and features. The hope in the end is to have an overarching tibble with all my combined data along with the smaller supporting tibbles.

Part 2: Data Transformation

• Initial Transformation of Data will involve the removal of some rows for the census data at the beginning and end, this being just random info that came with the data

• Missing Values: As previously mentioned, there is not any missing values to attend. If there is its data not relevant to my analysis and the columns themselves won’t be used

• Over/Under Field: I may be able to do this but I am unsure only because I also implemented the Libertarian party in my data. I am thinking that maybe I will need to have tibbles that have all the mentioned parties and one that focuses on Kamala and Trump by themselves

Part 3: Correlations

With the rework of my project, I have had to trash a lot of my previous corelations. I will give my thoughts on some potentially intriguing relationships

• Education vs Voter Choice: I believe you will find a solid relationship around counties education and where this leads voters to lean

• Poverty vs Voter Choice: I believe that poverty will have a relationship with voter choice for the respective county

• Unemployment vs Voter Choice: Lastly, I think unemployment will be a strong indicator as to voter tendencies in each county

Later on I hope to establish more relationships between polling and the census data, this hopefully showing a possible idea if things will remain the same this cycle or change.

Part 4: Modeling

Potential Models

Model 1: Simple Linear Regression

• The single variable for this model I would like to focus on is Bachelors’s Degree’s by county and relate this to prior elections and current polling

Model 2: Multiple Linear Regression (Continuous Variables)

• For this Model I want to look at Poverty and Unemployment combined to get an idea of its effect on polling and influence in prior elections

Model 3: Regression with a Categorical Variable

• For a categorical variable I will look at the counties themselves and try and look at the voting by gender and if there is a noticeable change between this election cycle polling and previous election results

Part 5: Analysis of Results

Due to my scrapping of my previous project, I don’t have any solid results to analyze. Looking at the data with a naked eye, I have identified various factors I believe will have a strong influence in regards to voter decisions in the past and potential voting decisions in the future.