Games with Models, April 22 Data


During her daily video statement today, Governor Gina Raimondo pronounced the “good news” that things seem to be leveling off in Rhode Island.  Based on my model, today was the peak for active cases.  (See here for the methodology.)

In the comments to yesterday’s post in this series, Mario pointed out that the state isn’t just updating its totals each day, but rather revising data going back.  So, the first chart in today’s post is a correction for yesterday:



Basically, the corrected data bumped the totals up a little and expanded out the downward slope.  Although my projected peak increased by five hospitalizations, the model still placed it today.

Happily, that did not happen.  The number of reported hospitalizations actually went down, rather than shooting up.  As you can see, hospitalizations went down substantially more than the day-to-day reports suggest because the state is now saying the past two days of reports should have been higher.  (Another note courtesy of Mario is that the hospitalization data is two days behind the other metrics.)



Compared with the numbers in yesterday’s post, the three key metrics current stand thus:

  • Cases:
    • Projection for 4/22: 5,731 (uncorrected)
    • Actual for 4/22: 5,841
    • Projection for 4/23: 6,083
  • Hospitalizations:
    • Projection for 4/22: 302 (uncorrected)
    • Actual for 4/22: 270
    • Projection for 4/23: 306
  • Deaths:
    • Projection for 4/22: 178 (uncorrected)
    • Actual for 4/22: 181
    • Projection for 4/23: 189

The corrected data suggests an important consideration.  For most of last week, the daily actual number of deaths was in double digits, up to 20 people.  The past three days have been single digit, which is a more-hopeful sign than the data above might suggest.

Another important note is that the big surge in cases I’ve been projecting for the day following each update is a function of the methodology.  It assumes that the ratio of the total number of people who have ever been hospitalized for COVID-19 in RI to the total number of people who have ever been diagnosed with it would stay pretty steady.  That is not looking accurate, inasmuch as the ratio has been falling pretty quickly — by about three percentage points per week.

To adjust for that, in the following chart I reduce the ratio each day by the compound daily growth rate for the past two weeks (i.e., how much the ratio has gone down each day on average for 14 days).  Under this scenario, yesterday would have been the hospitalization peak and tomorrow’s report will be 256.  I may run both charts for a bit to see how things go.


  • Mario

    Now that I have the discharge numbers, I can see the logic behind how the hospitalizations have been staying in the same range for so long, but no matter how I adjust things, I always show that things should be higher than they are. 290? We definitely seem to be near the peak of hospitalization usage regardless of what it is, but the case growth is still stubbornly high. I think tomorrow might be a good day, the legacy of that snowfall should be showing up around now, but I see no reason to think that the number of cases will start to fall overall: I’ll guess 257 for Thursday, but maybe the highest number yet on Friday.

    For the deaths (the only number I have ever felt any confidence in) 190. It actually ranges from 187 to 197, but I expect it to go up by 20 over the next two days, so I might as well split the difference.

  • ShannonEntropy

    You two are having a lot of fun with the numbers but there’s an old STEM-field saying: GIGO [ garbage in – garbage out ]

    The numbers presented by the gummint have been hopelessly politicized. There is now direct evidence that the overwhelming majority of infections are asymptomatic and the death rate is is in the 0.1 – 0.2% percent range. Just like every other seasonal flu bug IOW:

    The field of Medicine has been so politicized in the US that it’s sickening — no pun intended. Fer ex, did you know SMOKERS have a much lower infection rate than non-smokers and are much less likely to die if they do get infected ?? Chinese epidemiologists first noticed this back in January but somehow the American media & medical establishment never got around to reporting this

    Luckily I don’t get all my news from American sources; here’s a bit from the UK’s Daily Mail on the subject:

    In other news, CATS can get the Commie Cold a/k/a Chop Fluey. I have a 16 yr old indoor cat named Jewels and boy! is she unhappy about the new lock-down & social distancing rules I’ve had to impose upon her:

    • Mario

      That California data is almost certainly bad. NY just released the results of its own study suggesting that the infection rate in the city is 21.2%, putting the fatality rate at .837%. Their estimate could be completely wrong, but it’s probably wrong by something like 50%, not 500%. Maybe California just has a weaker strain of the virus, and I haven’t yet discarded the idea that Italian and Italian-adjacent people are more susceptible, but I think it’s more likely that people have seized on a study that tells them what they wanted to hear, and are ignoring ample evidence that the estimate just can’t work with the number of deaths coming out of New York.

      I saw the bit about smoking before. I think it could be true, but it doesn’t lead to a course of action.

      • ShannonEntropy

        The Boston Globe just reported that a third of randomly-tested people in Chelsea MA had C-19 antibodies:

        So that’s just more proof that my assertion is trve i.e. Coronavirus is just another seasonal flu bug and the mass hysteria resulting from the media coverage of it is worse than the disease itself

        My point about smoking is that you cannot trust anything coming out of the American media or medical establishment cuz everything they do & say is fatally tainted by a political agenda

        Do you think that if having had an abortion conferred the same level of protection that smoking appears to, that that fact wouldn’t have been headline-leading news in every American media outlet — other than Fox — and every
        American medical journal ??

        I think we both know the answer to that one

        • Mario

          If 1/3 Chelsea residents has had the virus, that would still point to a fatality rate of at least .3%, and it would mean that 60% of New Yorkers would have to have been infected already. More reasonable than the California data, at least.