Support for public health and safety measures that restrict civil liberties during the COVID-19 pandemic has been divided. Leaders calling for easing restrictions in the U.S by Easter became a flashpoint of concern. I was curious to see whether the rate of local confirmed cases increased, decreased, or stayed the same after Easter (April 12th, 2020).

The visualization below makes a strong case that the rate increased with the Easter protests.

 

 

Making this visualization involved studying the grammar of graphics (ggplot2, tidyverse, dplyr), and further developing of my skills at sourcing, compiling, cleaning and transforming data, along with programming some basic calculations to explore the relationships between variables. Multiple aesthetic elements, overlays and skills at understanding perceptual principles were used to polish the plot in order to communicate the findings intuitively.

Currently, I’m researching hospital utilization rate data sources, as increases in testing capacity add even more challenges to making valid inferences from the daily confirmed cases data. I would also like to get more involved in finding solutions to help officials make better informed policy decisions.

 

One way to help improve policy decision making on public health and safety is to establish Simple Random Sampling measures of both COVID-19 and SARS-CoV-2 voluntary testing at the municipal level. To date, I know of only one such effort: TRACE-COVID-19 - Oregon State University. I hope to have an opportunity to learn from the TRACE project team this summer and initiate discussions with local policy leaders ASAP.

 

Exploring the Doubling Rate Metric

A doubling rate (how many days it took to double the total number of confirmed cases) is a lagging indicator that is used below in an interactive Plotly graphic to visually explore the quantifiable projecting capability of our dynamic Daily Confirmed Case variable. This interactive second plot gives us a way to primitively explore when the doubling rate -as a projection- was accurate and when it was off. We can see that the doubling rate metric is rarely accurate at predicting from our dynamic variable. What’s most important is that the ‘doubling rate trend’ is improving (increasing), indicating a flattening of the curve, -except for one short reversal on May 10th.

 

 

More Advanced Visualization R Coding Forthcoming

Examining the doubling rates over the 14-day trend line with a confidence interval band may help to illustrate the imprecision. I can explore adding a more sophisticated interactive vertical confidence interval band for the doubling rate based on variance in the 14-day trend. Additionally I can look at possibly developing an algorithm that accounts for the effects of our dynamic explanatory variable; however, tinkering with the vertical confidence interval formula may accomplish improving the precision issue.