Games with Models, May 3 Data


The best news today comes from the number of positive cases of COVID-19 in Rhode Island.  The increase came very close to what my model was predicting, dropping the day-to-day growth to 2% and bringing the 14-day infection rate below one.  By way of reminder, that’s the threshold beneath which the disease is fading.

Of course, the number of total tests reported today was down, leaving open the possibility that the increase is only because the state didn’t do as much testing.  On the other hand, less testing could mean that fewer people needed to be tested, which is a positive.

As far as hospitalizations go, they continued to drop, although a little less precipitously than I projected.  The dotted line that I’ve added to the chart, by the way, represents the last day that the state was giving the public data on how many people were in the hospital because of COVID-19, rather than just testing positive for the disease while in the hospital for some other reason.



  • Cases:
    • Projection for 5/3: 9,451
    • Actual for 5/3: 9,477
    • Projection for 5/4: 9,633
  • Hospitalizations:
    • Projection for 5/3: 323
    • Actual for 5/3: 330
    • Projection for 5/4: 315
  • Deaths:
    • Projection for 5/3: 301
    • Actual for 5/3: 320 (note that this includes revisions going back multiple days)
    • Projection for 5/4: 325

  • Mario

    You case projection would imply a mere 6% positive rate if the testing stayed at only the same level as last Monday. Just applying today’s rate to last week’s number would give you an increase of 233. My numbers say 315, but there’s no real logic behind that except a slow steady decrease from the running average. I’d expect a minimum of 212.

    I tried to figure out the asymptomatic and pre-symptomatic population walking around unaware, just for fun, which I’d put at almost 20,000, down from a high of over 27,000 on the 22nd of April. From there you can figure it’s about an 82% chance that a group of ten would be safe from transmission. I don’t have any conclusions about this, except that if you were waiting until the safety hit some arbitrary threshold, I could estimate it. For instance, I would project it goes under 5% only on June 3rd, and I think we’re on track to see a little over 6% of the total population infected at some point. There’s no reason to think my numbers are right, mind you, but I don’t think they are wrong. What I really want to see is how things change as social distancing starts to fall apart. I don’t think it will be good, and I don’t think there will be much in the way of economic benefits to compensate, so I’m pessimistic overall.

    I think we’re on track for 536 total (official) deaths now. I know I once thought we might be able to limit it to 380, but now I think we’ll be coming right on that in just 8 days, and my numbers have been coming up extremely short on that side. I still see six tomorrow, and hospitalizations at 341 (+41, -25, -4), a little higher, and maybe higher the day after, but the trend remains negative.

    • Justin Katz

      I’m enjoying your explanations of your thinking, but to be clear, my intention was never to try to come up with the right number day to day. I set up the model and am simply reporting on what the numbers say each day, with just the one modification when it became clear my system wasn’t working when the hospitalizations started to turn the corner.