Covid-19-data: Data Issue: Michigan 06-05 Infect and Death Jump
Describe the issue:
- [ ] Incorrect number of cases or deaths
- [x] Suspicious number of cases or deaths
- [ ] Missing data for a locality
- [ ] Missing time-series information
- [ ] Other
Fuller details
Michigan State Infect & death has a big jump on 06-05. The government confirmed counts are 58,525 & 5,613 (https://www.michigan.gov/coronavirus/0,9753,7-406-98163_98173---,00.html) while nyt data is 63,744 & 5,854. is it due to the jump in probable cases?
guzhiling
All 12 comments
Hi @guzhiling on June 5, Michigan started reporting probable cases and deaths for the first state. We include those in our totals which is why the total increases so much.
albertsun
on 6 Jun 2020
Hi @guzhiling on June 5, Michigan started reporting probable cases and deaths for the first state. We include those in our totals which is why the total increases so much.
Thank you, Albert, for your timely reply. That solves my question. I want to confirm that right now there is no double-counting of probable cases from county level and state level, is it right?
guzhiling
on 6 Jun 2020
Yes, that's right. They are not double counted, but there are still some counties that report more probables than the state, in which case we use the number from the county instead of the number from the state.
albertsun
on 6 Jun 2020
I see. In that case, we use the county with the highest probable cases to represent the state-level probable cases, i.e. max(counties_i, state). That makes sense to me.
Thank you for your explanation and hard work!
guzhiling
on 6 Jun 2020
Cases in Oakland County specifically looks off, but perhaps they were just _really_ bad at confirming cases? 53% of all new cases in Michigan on June 5 are attributed to Oakland County alone. It's a jump of nearly 30% over that county's prior cumulative total, representing 0.2% of its population.
It's simply a huge outlier from all the other Michigan counties (and even its own increase in deaths).
mbauman
on 9 Jun 2020
@mbauman this is a limitation of the way this data is reported. Michigan added all their probable cases, including all those probable cases in Oakland county, on one day, even though many of those cases took place in the past. Our dataset primarily represents the total number of cases on any given day. Normally that would lead easily to a computation of new cases, but in this case, unfortunately, it doesn't.
albertsun
on 9 Jun 2020
Hi @guzhiling on June 5, Michigan started reporting probable cases and deaths for the first state. We include those in our totals which is why the total increases so much.
Are you using this backlog of data added on the 5th in your graphs on the main page to show which states are increasing vs decreasing? This is what it looks like and seems to misrepresent the data and trajectory of the state. I worry others may not dig in to the data to understand. Is my assumption correct, or am I misinterpreting the data? If this is the case it would be nice to add a footnote to the graphs when this change occurs. Thanks.
tgskiman
on 9 Jun 2020
The spike from one county on one day is making the overall data for the state rather useless. Please fix - I thought that was the whole point of this site with a day delay was to correct these sorts of issues.
westmich
on 9 Jun 2020
Michigan added all their probable cases, including all those probable cases in Oakland county, on one day, even though many of those cases took place in the past
Yes I fully understand that limitation. My point is that — even considering the temporal backlog — there is additionally a big _geographic_ outlier in where these cases occurred. But that very well may be true to the differences in how aggressively the counties are assigning probable cases.
I see you have nicely documented this in https://github.com/nytimes/covid-19-data/blob/master/PROBABLE-CASES-NOTE.md. Thanks for all your hard work in collecting this data!
mbauman
on 9 Jun 2020
@albertsun it looks like the bulk data was removed from the trend line which makes the charts read as would be expected. Thank you very much! The charts are much more representative of reality. Thanks again!
tgskiman
on 10 Jun 2020
Hi everyone, we're working on ways to consistently account for artificial spikes in the trends and how to label them clearly.
@mbauman I don't have a specific answer about Oakland County, but we've been looking at patterns with probable cases and deaths. They're often related to outbreaks in nursing homes and other facilities, especially earlier in the year when testing capacity was more limited. The share of probable cases in Oakland County is within the typical range we're seeing elsewhere, though it is toward the higher end.
lwaananenjones
on 10 Jun 2020
I’m not sure there’s a way around it e.g. some counties don’t report over the weekend causing a spike every Monday. we’ve just done 7-day trailing average to most our charts to smooth it for example the top 2 charts at the link below. Just my $0.02.
gmillikan
on 10 Jun 2020
Related issues
lsylvain
·
3Comments
jazz788
·
3Comments
desholmes
·
4Comments
bcunning
·
5Comments
dashdotdotdashdotdot
·
3Comments
Most helpful comment
Hi everyone, we're working on ways to consistently account for artificial spikes in the trends and how to label them clearly.
@mbauman I don't have a specific answer about Oakland County, but we've been looking at patterns with probable cases and deaths. They're often related to outbreaks in nursing homes and other facilities, especially earlier in the year when testing capacity was more limited. The share of probable cases in Oakland County is within the typical range we're seeing elsewhere, though it is toward the higher end.