An interactive simulation that demonstrates how important it is to keep COVID-19's reproduction number (R_t) below 1.

I think most folks understand how dangerous it would be if each infected person spread the virus to 2 more people, but I imagine many would be surprised at how quickly cases rise if each person with COVID "only" infects an average of 1.1 people.

The simulation also allows the user to choose a U.S. state to fill its current R_t estimate (sourced from R_t Live) and average daily cases, so they can see their own state's current outlook.

I adapted my simulation model from Rt Live's model as they lay out in their tutorial notebook. It accounts for the fact that it takes a few days for an infected person to pass on the virus by incorporating that delay (called the generation time) as a log-normal probability distribution with a mean of 4.7 days and standard deviation 2.9 days. Then, the number of newly infected people on a given day (y_t) is a sum that looks at all previous days, weighing the number of newly infected people i days ago (y_{t - i}​) by the generation time probability (g_i​) for that number of days and the effective reproduction number (which for the purposes of this simulation is being kept constant at R_e​). More succinctly,

Simulations run from today's date to four months in the future, and they treat every day prior to the start date as having a number of new infections equal to the "Initial daily infections" setting below the chart.

For the "Pick a U.S. state" dropdown, current R_t estimates and average daily cases are fetched directly from R_t Live, and they source their case count data from The COVID Tracking Project. These real-time R_t estimates have an inherent degree of uncertainty, so each estimate includes an 80% confidence interval, which essentially means that we can be 80% sure that the actual R_t value lies somewhere between those upper and lower bounds. You'll notice if you visit R_t Live's homepage that as you look further back in time, these confidence intervals become smaller with the benefit of more data.

These are the primary tools I used in this project:

• Data visualization: Vega-Lite
• Statistics package: jStat
• Data fetching: SWR
• Data wrangling: Data-Forge
• Math typesetting: KaTeX via react-katex
• Web framework: Next.js
• Web design library: Bootstrap via React Bootstrap

Also, I have to say, while the focus of this project was the data visualization, not the web development, this project contains the most beautiful React code I've ever written. ⚛️