good work, Thanks

Thank you for the good local journalism.

Something that concerns me is the lack of availabie testing.  The cost of the test is a bottleneck, I have heard estimates of ~$3,000 for a test.  From what I have read, many positive tests were from folks wih no discernable symptoms.  Some doctors have told patients 

that the test is only for those presenting severe symptoms such as fever.  This advice defies the number of positives from those with no symptoms.  Possible exposure is not a valid reason to test according to some doctors.  Set up a drive through test center with free tests and we may have far different results.

Would be nice if the windshield and rear windows were cleaned as we move forward.

Wear your mask if you can find them, or make your own.  Covid-19 has shown what an exceptional health delivery system we have.  Go long on the profit centers and private equity grifters.

I agree with Nate Silver’s expert-at-stats crew, that the data & predictions like these are, regretably, far too incomplete for confidence in this effort. Why? Please read https://fivethirtyeight.com/features/coronavirus-case-counts-are-meaningless/.

Hi Wynne Lee,

Remember, as stated in the articles, the forecasts are for confirmed cases and not for total infections. As such, I believe they are reasonably “complete” for confirmed cases because their has been no indication of increased (or, perhaps lower) rates of testing; neither is their any evidence to suggest that testing accuracy has changed.  So far, the assumption of the testing rates and their accuracy being relatively constant seems to be holding. Monitoring this is important, however,  and if changes are reported, it is possible to build these changes retrospectively into the baseline and update as well as use this information in forecasts down the road.

What we lack  is good information on the ratio of positive tests to infections and this is unlikely to be avaiable in the time to the expected surge peak (and more likely, far beyond it). Without this and the lack of universal testing, like you I would lack onfidence in any forecast of total infections. However, if we  had rasonable information on the ratio of confirmed cases to total infections, then the consequences of the confirmed cases such as hospitalizations, deaths, and recoveries could be readily extended to all infections.

Clearly, confirmed cases understate  toral infections at this point. As such, they can be viewed as providing a conservative portrait relative  to the virtually non-existent information we have on all infections, From this standpoint alone, I believe that are far superior to both doing nothing and simply guessing about what lies ahead - they can be viewed as providing a “best case” scenario of what lies ahead.

Thanks for your work.

Where can I find the raw numbers of confirmed cases/number of deaths in Whatcom County since the beginning of the outbreak?  I can find the latest stats at the Whatcom County Dept. of Health, but looking for the data of each day - hopefully in one place.

Hi David MacLeod,

For all of its faults (e.g., slow reporting and then a “data dump”) a consistent source is found at Johns Hopkins University (JHU). For “today’s” data you can use the map.  Enlarge it so that you can idenify Whatcom County and then point and click on the
red circle hovering over it. Once open it will state that it is for Whatcom County and show the cumulative number of confirmed cases and deaths etc. up to “today” (my advice is to wait until well into the evening to use this resource).

To get “today’s” data from the map, go to:   coronavirus.jhu.edu/map/html

For the data prior to “today” you need to go to JHU’s archived data on GItHub. The files are organized by date, one file for each day. The archived data files prior to March 24th are really messy so my advice is to start on the 24th if you want to avoid a whole lot of frustration. Once you open an archived file you will find the data organized alphabetically by county. Because many county names are found in more than one state (which is not the case for Whatcom County) you will see the name of the state (along with FIPS code, latitude and longitude of the county seat) and confirmed cases and deaths, etc.. Sometimes when  you open an archived file, the right side is obscured so you have to go to the bottom (a long way down) to reach the toolbar to push the file open a bit more to the right in order to see confirmed cumulative cases and cumulaitve deaths.

Once you have the URL for an archived file for a given date, you can to to the other archived files by changing the date followed by pressing the “enter” key.  Here is an example. Go to the URL for the March 24th archived file: https://github.com/CSSEGISandData/COVID-19/blob/master/csse_covid_19_data/csse_covid_19_daily_reports/03-24-2020.csv  Once you have Whatcom’s data for the 24th, change “03-23” to “03-25” to get data for March 25th and so on.

Once you have set up the archived data so it is current, you do not need to return to the GitHub site. All you need to do is update daily from the JHU map.

Good luck.

Thank you David for the detailed and helpful directions.  I’ve been able to get data from March 29 onward from the Whatcom County website https://www.whatcomcounty.us/3329/Novel-Coronavirus-COVID-19, but your link helped fill in some gaps before that.  The data doesn’t align 100% between both sites, but basically seems to just be off by one day in some cases (reflecting time of collection probably). 

Right now it looks like the rate of confirmed cases is doubling every 8 days, while the number who have died is doubling every 5 days.  The next few days will be interesting to see how much we’ve managed to flatten the curve. 

Hi David, MacLeod,

Great to learn that you got what you needed. Your comment about differences between the reported data found at the Whatcom website and the JHU website being “almost in alignment” is not surprising. Because I was pretty certain that I would be doing more than Whatcom when I started this effort, the JHU site was clearly the best for me - “One stop shopping.” Once the data are posted there, however, it is apparent that nothing is retroactively corrected (The staff  must be overwhelmend with their worldwide collection effort. How could they have time to do this?). As one of the two cases I have encountered so far,  for Kittitas County, the JHU archived files report 5 confirmed (cumulative) cases on March 24th, 18 confirmed (cumulative) cases on the 25th and then 8 confirmed  (cumulative) cases on the 26th. Clearly, the report for March 25th is an error. Instead of carrying this error, however, I estimated the number for the 25th via interpolating between the 24th and 26th and noted in my Kittitas file that it is interpolated (with the excel command that calcualted the daily geometric rate of change next to the note).  Better the estimate (and so noted as such in the file) than leaving the erroneous number.  Be prepared.

Excellent ‘flying blind’ article that reports some of the challenges of modeling projections, using WA data. Cliff Mass explains the crucial ‘initial conditions’ problem, how to improve those estimations, plus how projections for WA are changing over time. Models shown mostly include error bars, thank goodness, as those are as (maybe more) crucial as center lines for communicating about what the models suggest. https://cliffmass.blogspot.com/2020/04/flying-blind-on-coronavirus-why-random.html

It’s also very worthwhile reading the comments to Mass’s flying blind article.