Since the correlation among the factors in the Coronavirus Disease Data is very high and meanwhile considering the influence of collinearity, only one factor will be chosen from the Coronavirus Disease Data and added into the two-way ANOVA model. Based on the One-way ANOVA of each feature in the Coronavirus Disease Data, the effect of the 7-day average of count on total percent in a day is most notable. For the ANOVA of the average tip amount in a day, the case count will be added into the analysis.
1. Total payment per day
The first ANOVA model is conducted to compare the main effect of the 7-day average of count from the Coronavirus Disease Data and Public Health Policy Data on the total payment per day. For the 7-day average of count, it yields an effect size with ηp2 = 0.82 (95% CI: 0.79-0.84), indicating 82% of the variance in the total payment per day can be explained by pandemic data (F(5, 327) = 846.3, p < 2e-16). And the 14% (95% CI: 0.07-0.20) of the variance in the total payment per day can be explained by the public health policy (F(2, 327) = 846.3, p < 2e-16).
Figure 1 illustrates the difference of Estimated Marginal Means (EMM) of the sum of total payment per day under different levels of a 7-day average of count. Because of the outbreak of the COVID-19, the EMM of the sum of total payment per day decreased from $4,015, 409.05 to $1,116,858.25. The fluctuated and continuous drop is shown as the 7-day average of count increased from (1,271] to (3820,5782], and finally hit the bottom, with $317,534.0. Figure 2 indicates the EMM of the sum of total payment per day diminishes as the impact area of public health policy expands. When there is no policy, the EMM of the sum of total payment per day was $3,880,781.85. After the policy associated with COVID-19 was announced, the EMM of the sum of total payment per day falls to $627,99.39 and down to the bottom after the ‘NYS on Pause’, with $147,564.52.
Figure 3 presents the tendency of total payment per day in a day during the COVID-19. It reveals an obvious drop along with the increase in as the new confirmed cases and the whole of New York State is paused meanwhile, from 4.0 million dollars drop to 0.14 million. As phase 1 of reopening started, the total payment per day begins to resume gradually, hitting a high of 1.86 million in November. However, the second outbreak of the COVID-19 resists its slow recovery and the total payment per day shrinks to 0.3 million in December.
2. Average tip amount in a day
A two-way ANOVA is performed to analyse the effect of the Coronavirus Disease Data and Public Health Policy on the average tips in a day. The result demonstrates that there is a statistically-significant difference in the average tips under different degrees of the case count (f(4)=395.8, p < 2e-16) and distinct levels of public health policies (f(2)=267.3, p < 2e-16). Additionally, both of case count and public health policies have a high effect size, with respectively ηp2(case count) = 0.39 (95% CI: 0.31-0.46) and ηp2(public health policy) = 0.40 (95% CI: 0.32-0.46).
Figure 4 discloses the impact of the pandemic on average tips. The EMM of average tips before the outbreak of pandemic is $2.15, while as the daily new cases increases to (1, 200], the EMM of average tips drops to $1.93. The EMM of average tip increases to $1.87, even the daily new cases increase to (573,860], however, it finally hit the bottom with $1.72 when the daily new cases in (860,6353]. Figure 5 detects the change in average tips under three levels of policies, and it clearly indicates the reduction of average tips from $2.18 to $1.55 as the policy change from 0 to 2.
Figure 6 indicates the trend of tip amount under the impact from COVID-19 and public health policy. The average tip amount drops from $2.3 to $1.3 during the period between 5 March 2020 and 4 April 2020. After phase 1 of reopening, the average tip reveals a steady increase to around $2.0. However, as the second wave of the epidemic spread, average tip begins to decline slightly.
3. Change of demand between 2019 and 2020, by hour and day of the week
The demand for taxis between February to December in both 2019 and 2020 are chosen to be analysed. The data exposes that the total demand in 2019 is 79,068,996, while the value falls to 17,628,388, with an 77.7% reduction. With respect to displacement in relative demand by the hour and day of the week, from Figure 7, the fall in demand is slighter during the working hour, 6:00-17:00 from Monday to Friday, with -0.74 on average. And the decrease in demand at weekend from 6:00 to 17:00 is around -0.77. For the late-night, from 19:00 to 5:00, the average decrease in demand is -0.83.