Data for Massachusetts displaying weird

aatishb / covidtrends

Tracking the growth of COVID-19 Cases worldwide

https://aatishb.com/covidtrends/

MIT License

301 stars 107 forks source link

Data for Massachusetts displaying weird #213

Open stephen304 opened 4 years ago

stephen304 commented 4 years ago

Looking at the data, it seems normal (all around 120k), so maybe this is just a bug in the graph. It looks like the line is going to negative infinity (I can't find the end) @aatishb

drlaurenwasson commented 4 years ago

Came here to leave the same comment. Thanks for beating me to it!

DesiOtaku commented 4 years ago

I tried to see what could be causing the bug but I can't seem to find it.

Only lead I have so far is that it is downloading the data directly from https://raw.githubusercontent.com/nytimes/covid-19-data/master/us-states.csv and the Massachusetts data looks fine there.

stephen304 commented 4 years ago

The data looks extra weird today, it looks like the cumulative confirmed cases must have decreased in the data, as Massachusetts moved left on the x axis:

@aatishb

aatishb commented 4 years ago

Hi all, Thanks for filing and looking into this. It looks like Massachusettes changed the way their data is being reported which resulted in fewer case counts. It's a bit easier to see on the linear scale view.

https://github.com/nytimes/covid-19-data/issues/447 https://www.nytimes.com/interactive/2020/us/massachusetts-coronavirus-cases.html#anomaly-notes https://www.mass.gov/doc/covid-19-dashboard-september-2-2020/download

Since it's a data source / methodology issue we don't really have a great way of dealing with this on our end. We can wait and see if NYT decides to retroactively update their data to account for the methodology.

stephen304 commented 4 years ago

@aatishb Does that also explain the graph not showing any points for 9/2-9/8? There should be data for those days but no points exist on the graph

Edit: Just realized that I was looking at the wrong number in the data - I guess the negative change in the data is what caused the missing points

aatishb commented 4 years ago

@stephen304 Exactly. It seems the large negative change in numbers on 9/2 led to the 'change in the previous week' as being negative for the entire next week (9/2 - 9/8).

A potential workaround might be to drop negative values before graphing, that way the graph would instead show a jump from 9/1 to 9/9. However if this is the only place this is occurring it might be worth waiting to see if they update the data source first.

vjandrea commented 3 years ago

Hello, the graph looks bizarre for France:

and Spain shows a drop: