MODERATOR NOTE: coronavirus resources & updates: https://wolfr.am/coronavirus
-- you have earned Featured Contributor Badge
Your exceptional post has been selected for our editorial column Staff Picks http://wolfr.am/StaffPicks and Your Profile is now distinguished by a Featured Contributor Badge and is displayed on the Featured Contributor Board. Thank you, keep it coming, and consider contributing to the The Notebook Archive!
I thought this was a fantastic post and I was very much interested in looking at the daily trends in testing data.
Unfortunately, I found that there is a great deal of noise in the state level testing reports. Some states are reporting 100% positive rates in the last day.
Import["https://covidtracking.com/api/states/daily", "RawJSON"] /.
Null -> Missing // Dataset;statesDailyTracking[Select[#state == "MD" &]][1 ;; 5])
I do not know how to paste my output here.
However, it is clear that the way that testing is being performed unevenly across the country by state (and perhaps by county within state) makes the analysis of serial testing data much more complicated than just looking at the rolled up countrywide results.
Nothing about this data is easy, but live data almost always has difficulties.
It may also be the way they are curating the data. Since it is cumulative, if they cannot find any report by the state of negative results, they may just be taking the last number. For Maryland it looks like the cumulative negatives have been stuck at 94 for quite a while. By now there are so many tests being done, that they may not be getting good numbers on the negatives.
I played around a little with the state by state data yesterday using the positives, which seem to tracking with other data sets well. You might be interested in the attached notebook. The graphic in it is explained in this post: Predicting Coronavirus Epidemic United States
They are updating the data early each day so you can get up to date results, without waiting until the next day..
I have played around a little with the state data and I looked through your notebook closely to see if I could figure out a way to enhance the analysis to identify clusters of states that are testing a lot and clusters that are testing only those with obvious symptoms so that I could then estimate a sequence of percentage positive tests over time and see if they start to converge.
Then the absolute counts of positives could be used to estimate if we had turned a corner, at least in those jurisdictions with sufficient testing.
But the data are too thin and there just does not seem to be any state that is testing as much as NY. Maybe in a few days.
The surge in testing may not produce much. The news yesterday was that over 300,000 tests have been performed in the U.S., most over the last week, but no state is showing an unexpected surge in cases. In fact NY is starting to look like the quarantine is working. Most likely good epidemiology will test the right people without a lot of extra tests. Epidemiological methods have worked long before tests were available. Not much information is available about the test. But it also may be that it would be better to test a contact after the incubation period as there may not be many virus particles in secretions until the virus causes cells to break down and release the new virus particles.