Bay Area Entrepreneurs in Statistics

The Remdesivir paradox modeled. The World Health Organization Solidarity study concludes there is no benefit to Remdesivir for severe Covid-19 cases (ref: https://www.sciencemag.org//remdesivir-and-interferon-fall) while FDA has approved its use (ref: https://www.fda.gov//press-announcements/fda-approves-firs). Timing is critical and so I wrote a program to model various scenarios as visualization aids. The attached image displays one scenario: the blue curve is the vir...al count and red curve antibodies with no Remdesiver treatment. The vertical orange and purple lines are the times after infection for the initiation of Remdesiver. The corresponding orange and purple curves represent the viral loads for these two different timings of Remdesiver. I am still working on the details of what happens to the antibody titer when the viral load decreases, so it's a work in progress. Comments accepted.

Did you know you can calculate the odds of at least one person carrying the Coronavirus in a crowd using your phone calculator? Most phones have a basic calculator app. Use the equation at the end of this post and plug in your own numbers. A screenshot of example calculation on the phone calculator attached. You can find current case positives and other Covid19 data at https://covidtracking.com Inputs: ... (C) current number of case positives in a given area, like a county, State, or country; (B) bias factor converting daily case positives to actively infectious population (a fudge factor to account for individuals not tested, individuals infections prior to and post testing; (P) total population in area; and (S) size of the crowd Output: (R) % risk of infected individual in the crowd Equation: R = 100*(1-(1-C*B/P)^S) Example: Contra Costa County currently has C = 120 (rolling ave); B = 10 (i.e., for every person tested positive today there are 10 who are currently infectious); P = 1.2*10^6; S=50 (a big party!) Calculation: Risk = 100*(1-(1-120*10/1,200,000)^50) = 5% If B = 20, Risk = 100*(1-(1-120*20/1,200000)^50) = 9.5% If S = 100 Risk = 18% Bottom line: while a 5% risk doesn't sound like much, just attend a few more parties and the risk will increase dramatically. So wear a mask, keep your distance, wash your hands, avoid those large parties for now.

What is the risk of someone in the crowd positive for Covid-19? It depends on the size of the crowd and the infection rate in the area. Here is a risk calculator that can give you a sense of the numbers: https://covid19risk.biosci.gatech.edu Note: this is not the same as what is your risk of infection.

U.S. Covid-19 cases going up and mortality going down (slightly). This, to me, means the medical profession and hospitals are doing a good job keeping us from killing ourselves when we drop our guard and stop doing what is necessary to slow the rate of new cases. I leave two charts that show the differences for the first and second half of the first 250 days of the pandemic. The first chart is for new cases and the second for daily mortality. In each chart the red lin...e is the full data set starting Mar 4, 2020 with values ranked from low to high plotted against the probability of occurrence. The blue line is the first half of the data set, ranked by magnitude and the brown curve is the second half. New cases in the second half exceed overall new cases in the first half, while daily mortality is the reverse, though the numbers are still high at > 1000/day.

What are the odds? Let's say you are a supervisor running a process control for keeping a discharge under control. The plant manager comes to you and says "hey our discharge limit for mercury is 20 nanograms per liter and we've been running an average of 15 with no discharge violations for the past year. Loosen up! We can save $100,000 a year if we let our average go to 16 ng/L. So would you do it? To determine your response you go to your on-staff statistician (who al...most got the budget axe but managed to weather that one). He asks for the relevant numbers: current average = 15, standard deviation = 2, discharge limit = 20, requested new average = 16, new estimated standard deviation = 2.1 (assuming constant RSD). The results? Expected violations at current average and standard deviation < 1%. Expected violations at new set point = 4%. Each violation is $30,000. So do you change the process? Oh, and to work with your manager, who is not a numbers type, this chart might help. The blue curve shows current operations; the green curve shows projections under the proposed operation; and the vertical red line is the discharge limit.

On September 9, 2020 many of us woke to dark skies burning red but with local air that was not smokey. The explanation was a cool marine layer of fog between us and a very deep warmer layer of smoke that blocked the sunlight. That two such layers did not mix may be surprising to some, but a simple kitchen experiment can demonstrate the phenomena. Using food coloring dyes, add a couple of drops of one color dye to some cool water. Then warm some water in a separate container and add a different color dye. Now, carefully pour the warm color water down the side of the container with the cooler water on the bottom. Watch what happens. Spoiler attached.