Project: US State Vaccinations and Political Affiliations

In the past year, the world came to a standstill. The COVID-19 virus has affected so many lives and is still present in society. However, we are moving back towards what we would consider "normal life". The COVID-19 vaccine has become more widely available in the US and restrictions are starting to be lifted. Data science students became interested in looking at vaccination data sets. Questions that arose from this is if vaccination rates are different in certain states due to political affiliations, like whether the state leans red (Republican/conservative) or blue (Democrat/liberal). That is what we aim to determine in this first project.

import pandas as pd state_election = pd.read_csv('Election Dataset.csv', dtype={'RESULT':'category'}) # converts the csv file to a pandas dataframe state_election.head() #prints the first 5 rows

vaccine_data = pd.read_csv('us_state_vaccinations.csv',parse_dates=['date']).dropna() vaccine_data['location'] = vaccine_data['location'].str.upper() # takes the columnn location and converts each value uppercase vaccine_data.tail(10) # displays the last five rows

def search(state,column): global state_name state_name = state.upper() # ensures that no upper case and lower case error occurs search_series = vaccine_data.loc[vaccine_data['location'] == state_name, column] # we use the index attribute of series to set the index of the series created search_series.index = vaccine_data.loc[vaccine_data['location'] == state_name, 'date'] return search_series search_series = search('wyoming', 'people_vaccinated')

import matplotlib.pyplot as plt import seaborn as sns def lineplot(line_series): sns.set_style('darkgrid') # changing the background to a darker color sns.relplot(x = line_series.index, y = line_series, data = line_series, kind = 'line', color='r') plt.xlabel('Date') plt.ylabel(line_series.name.replace('_',' ').capitalize()) plt.xticks(rotation=45) # to ensure that all dates are seen clearly plt.title(label=line_series.name.replace('_',' ').capitalize() + ' in '+ state_name.capitalize() + ' by Date') return plt.show() lineplot(search_series)

import numpy as np from scipy.optimize import curve_fit def log_model(x, β0, β1, β2 ): return β0/(1 + np.exp(β1*(-x+β2))) def beta_solving(series): x_model = list(range(len(series.index))) # calculating domain of graph y_model = series try: initial_guess = [series.iloc[-1], 1, len(series.index)/2] found_betas, covariance = curve_fit( log_model, x_model, y_model, p0=initial_guess ) except: found_betas = [np.nan, np.nan, np.nan] # will return null valus if the curve fit produces no beta value return found_betas def fitting_model(fit_series): x_model = list(range(len(fit_series.index))) y_model = fit_series β0, β1, β2 = beta_solving(fit_series) # calculating the beta values for curve fitting fit_model = lambda x: log_model( x, β0, β1, β2 ) sns.relplot(x = x_model, y = y_model, marker = 'v', color='b', kind='scatter' ) sample_xs = np.linspace(0, 120, 100 ) # producing 100 values equally spaced between 0 and 120 as sample x values plt.plot( sample_xs, fit_model( sample_xs ), color = 'r' ) plt.xlabel('Number of Days') plt.ylabel(fit_series.name.replace('_',' ').capitalize()) plt.title(label=fit_series.name.replace('_',' ').capitalize() + ' in '+ state_name.capitalize() + ' by Days') return plt.show() fitting_model(search_series)

state_summary = {} for state in state_election['State']: betas = list(beta_solving(search(state,'people_vaccinated_per_hundred'))) #nested function state_summary[state] = betas state_summary = pd.DataFrame(state_summary).transpose() state_summary.reset_index(inplace=True) state_summary.columns = ['State','Maximum number of vaccinations', 'Rate of vaccination', 'Time of maximum increase'] state_summary = state_summary.merge(state_election, on = 'State').set_index('Abbreviation') state_summary

sns.heatmap(state_summary.corr(), annot=True ) # annot allows you display the corr coeff value plt.show()

from scipy import stats blue_sample = state_summary[state_summary['RESULT'] == 'BLUE'].dropna() red_sample = state_summary[state_summary['RESULT'] == 'RED'].dropna() def same_mean_hypothesis(column,α = 0.05): t_statistics, p_value = stats.ttest_ind(blue_sample[column],red_sample[column],equal_var=False ) reject_H0 = p_value < α if reject_H0 == True: answer = f'The mean {column.lower()} are significantly different for both political parties as p value of {p_value:.2E} is less than alpha level of {α}.' else: answer = f'The mean {column.lower()} are not significantly different for both political parties as p value of {p_value:.2E} is greater than alpha level of {α}.' return answer same_mean_hypothesis('Maximum number of vaccinations')

same_mean_hypothesis('Time of maximum increase')

(blue_sample['Maximum number of vaccinations'].mean(), red_sample['Maximum number of vaccinations'].mean(), blue_sample['Time of maximum increase'].mean(), red_sample['Time of maximum increase'].mean())

Throughout the course of our first project we utilized data science tools/packages to analyze data sets dealing with COVID-19 vaccination rates in the US and each state's political affiliation (based on the 2016 presidential election). The analysis helped us answer our question of whether political affiliations have anything to do with vaccination rates. From our specific data sets we concluded yes there is a correlation or relationship between these two variables. Hypothesis tests we performed concluded that blue states will tend to have significantly higher max number of vaccinations long term over red states. This was also seen as the case for the time of maximum increase variable. However, the max number of vaccinations long term is our most important variable to look at to answer our question of interest. These results rather make sense since most red states were more lenient with restrictions and have not reacted well to the COVID-19 pandemic in general. One thing to note is this is just one data set and probably an over generalization, but definitely is important to think about either way.