Data Analysis of the COVID-19 Pandemic in New Jersey
GitHub | Shiny App | LinkedIn
The skills the author demonstrated here can be learned through taking Data Science with Machine Learning bootcamp with NYC Data Science Academy.
Introduction
Data shows that the Covid-19 Pandemic has changed the world as we know it. For over two years, biologists, data scientists, politicians, and many others have been working to contain the virus and stop or reduce the spread. These efforts have come in the form of lockdowns, mask mandates, vaccines, extensive research, and more.
Some of this research has included demographic analyses of different regions and cities to determine whether a correlation exists between virus spread and certain demographic statistics. To further study these potential relationships, I decided to perform an analysis on my home state of New Jersey, using Exploratory Data Analysis (EDA) techniques in R and using Shiny to display my findings and create an interactive app in which the user can examine these potential relationships for themselves.
The Data
To begin, I obtained New Jersey Covid-19 time series data on both the state and county levels from the Act Now Coalition. Data collection began on March 1, 2020 and has been updated daily ever since. When I obtained the files, they contained data up through March 1, 2022, resulting in exactly two years worth of Covid-19 data to analyze.
The data sets contained more variables than I could ever have needed, ranging from test positivity ratios and case densities, to vaccination rates, to potential risk levels, and much more. In order to best answer the questions that I had sought to investigate, I narrowed it down to six Covid-19 statistics: Infection Rate, Percent Initiated Vaccination, Percent Fully Vaccinated, Percent Fully Vaccinated and Boosted, Total Cases per 1,000 People, and Total Deaths per 1,000 People.
Next I found a demographic data set from Kaggle, which contained New Jersey demographic data from the 2020 Census. Much like the previous data sets, there was more demographic information in this data set than I could use for one project, so again I had to narrow down my selection. I again chose six statistics: Percent High School Graduate, Percent of Adults 25+ with a Bachelor's Degree, Percent with No Health Insurance, Median Household Income, Percent Living in Poverty, and Population Density.
Data Analysis
To determine the correlation between the different demographic statistics and Covid-19 statistics, I created interactive scatterplots with the demographic statistics on the x-axis and Covid-19 statistics on the y-axis. I also had the correlation coefficient print above the graph, so that the user could see the potential relationship between the variables graphically, while also have a metric provided to them that would tell them how strong the linear relationship is. In total, there are 36 different scatterplots that the user can create between the different variables.
Very Strong Correlation
Out of all of the scatterplots, the strongest linear correlation was between percent of the population with a Covid booster shot vs percent of adults over 25 with a Bachelor's Degree. From looking at the graph, it is clear to see that there is a positive linear correlation between the two variables. The Pearson correlation coefficient is 0.904, indicating a very strong positive correlation.
Percent of adults with a Bachelor's Degree was strongly correlated with percent of the population that had initiated vaccination and also with percent fully vaccinated, but the correlation becomes even stronger when it comes to the booster. Since it is most likely not the case that having a Bachelor's Degree is what directly causes someone to get a Covid booster shot, we can infer that there are other underlying factors that lead to this correlation.
Strong Correlation
Although not quite as strong as the previous example, there is a substantial linear correlation between total Covid deaths and percent of the population with no health insurance. From the graph above, we can see that as the percent with no health insurance increases, the death rate per 1,000 people increases as well. The correlation coefficient is 0.779, denoting a strong positive linear relationship between the variables.
There was also a strong positive linear relationship between total cases and percent with no health insurance; across many different demographic statistics, when there was a linear relationship with total cases, that relationship almost always got stronger when that same demographic statistic was paired with total deaths, and percent with no health insurance was no exception. Although we don't have proof that not having health insurance is the direct cause for higher death rates, there is certainly room for speculation. There are most likely other underlying factors that contribute to this correlation as well.
Weak Correlation
Across all of the demographic statistics, there was one that did not have any strong correlations with any of the Covid statistics: population density. For example, when we plot total cases vs. population density, as pictured above, there is no substantial linear trend. The correlation coefficient is 0.282, indicating a weak positive linear relationship. It is worth noting that Hudson County is a bit of an outlier; its population density is more than twice as large as the next highest county, which could be causing it to weaken a linear relationship that might have otherwise been stronger if not for Hudson County's extremely high population density.
Data Science Summary
The table above shows the correlation matrix between all of the demographic and Covid statistics, as well as highlighting the strength of each linear correlation. As you can see, the strongest correlations are with Covid booster shot statistics. In general, Covid-19 case and death rates increase as the percent of the population with no health insurance increases, whereas case and death rates decrease as median household income increases. Likewise, death rates increase as the percent of the population living in poverty increases, and decrease as the percent of adults who have graduated high school increases.
Additionally, as the percent of adults who have a Bachelor's Degree increases, there is an increase in the percent of the population who has initiated vaccination, completed vaccination, and received a booster shot. The percent of the population who have received a booster shot increases as the percent of adults who have graduated high school increases and median household income increases, and decreases as the percent of the population living in poverty increases. Population density and infection rate have no strong correlation with any of the other statistics.
Further Analysis
If given more time and more data, I would continue to study the connection between demographics and Covid statistics by analyzing case, death, and vaccination rates by race and political affiliation, comparing vaccination rates to case and death rates, and studying counties in other states or on the national level.
About Author
