Zombies in Spain? about data

[owahltinez]

I see it too. Here's what's happening: Spain has stopped reporting historical data, so on the 24-25 of May the data for Spain starts coming from a different source (ECDC) which has different baseline numbers.

It's tricky because the ECDC data source does not have regional data, test counts and hospitalization information, so we can't just remove the old data source. Further, ECDC has clearly erroneous data for Spain in March. I'll try to setup the data pipeline to ignore national deceased counts from the official source and get them only from the ECDC, maybe those are a little more consistent.

[owahltinez]

Unfortunately, even the ECDC data appears to have the drop of ~2K in deaths. It must be an adjustment in the way the deaths were being counted (it would be the second time a big adjustment is made by this reporting country, the first one was in April for the way they counted positive tests).

I don't think there's a good way to deal with this problem. You either take the drop as the true value, or you adjust the data to keep it monotonically increasing but deviating from the official counts. Adjustments in the way metrics are computed are not uncommon, but they typically occur in count of confirmed cases and the change is smaller than 7%.

This is why we now provide both the "new_deceased" and "total_deceased" columns. If you want to ignore bogus values, use the "new_deceased" and skip negative values. If you want to keep the graphs as close to the official reporting as possible, use "total_deceased" but keep in mind that adjustments can happen.

[jmullo]

Ok, thanks.

[owahltinez]

Closing for now, feel free to reopen if there's something else we need to do on our end