City data is gone (state/province level only now)

[jkirkland-cityoflewisville-com]

Around 21:30 CST the "Province_State" field's data starting showing complete US state names instead of [City,State Abbreviation]. Hopefully the city/county level data will come back soon

[jrmatchett]

Bummer! Although, I understand it's probably a very labor-intensive process to compile county-specific data as the situation escalates.

[Hammerfest]

Bummer! Although, I understand it's probably a very labor-intensive process to compile county-specific data as the situation escalates.

this may be true, but thats no reason to remove existing data... (I know your not in charge, just a reply for those reading)

---

Ya, there removal of this data is disturbing and unhelpful

[bemosior]

It's very helpful, for educational purposes, to observe patterns of spread at the city/county level. For state-level data the NY Times dashboard is fine, but I've been coming here for the granularity.

I'm guessing here, but it seems more likely that it is an accidental configuration issue, as even as recently as a few hours ago data was being added (https://github.com/CSSEGISandData/COVID-19/commit/473681f6be0ccbba896557f1649611f46ed8b891) to this repo that includes county-level granularity.

These folks are working very hard to share this information with us, so let's be patient and see what they have to say tomorrow.

[mw32]

What I'm curious is how the data is pulled. If these are all automated or manual somehow (someone pulling strings behind the scene). I really hope it's just a config error, which can be easily fixed, but without knowing what goes under the hood, hard to say. I came here because just a couple hours ago max (~ 7 PM PT) I saw county data, but upon refreshing just 20 min ago they are gone. Then I stumbled upon this problem tix, and I can see that last update hrs ago still has county data (which was way before last time I saw the dashboard had them) so here's hoping it's a config change.

[JeffLoucksPersonal]

Canada now Geotagging at the provincial level only, city data is gone? This is a big loss of transparency. We need it back. You can imagine it is a very different thing to have a geotag 2000 KM away from the actual location as it is in Ontario or Quebec,

[mw32]

I think the problem should say "Some granular data is not displayed". At least for me, California counties are no longer displayed but I can confirm they are still there.

[jrmatchett]

The hosted feature layer was definitely changed to include only state-level data. How exactly are you confirming that the county-level data are still there?

[mw32]

Check out the last update (4-5 hrs ago) to the time series data: https://github.com/CSSEGISandData/COVID-19/commit/473681f6be0ccbba896557f1649611f46ed8b891 That still has counties in California, at least the county I was looking for.

[urbnfntsys1973]

County data seems to still be in the auto data files. Likely a decision to combine at a higher level.

[jrmatchett]

The time-series data have been compiled at 00:00 UTC each day. We'll see tomorrow if the next date period still includes the county-level data. Either way, that's not the data source that is used to render points on the map.

[ciscorucinski]

This change might be related to my suggestion in ... #284 (Mechanism to adjust US Covid cases by different groups)

I suggested a "mechanism" to change the behavior of the grouping. This could all happen locally on our computers. This would allow people to view the USA cases on the county, state, or country level.

[mw32]

I hope it's not a decision to combine all the data at that high a level because granularity matters.

I like the idea of grouping at higher level but ONLY when the view is at the bigger level. When you zoom in it should still display granularity at the level you want to see (county is good enough for me). In other words these grouping should be dynamic at the user interface, which is not the case here (I zoom in all the way to bay area and still only see California).

[jrmatchett]

That'd be ideal, but it also places a burden on the data maintainers to ensure accuracy at the finer scales. I've already noticed a few counties around me that have had incorrect data for several days.

[mw32]

Sure, same here. Alameda showed only 2 earlier (before it disappeared) when there's already 3. That's minor and much better than not having anything. That goes back to my earlier question - how is the data acquired and compiled. Anyway I hope I can see the granular data again.

[ciscorucinski]

@mw32 I agree with you partially. I do like the current State-level grouping as default, but I want to still see the other levels (county and country).

Zooming in shouldn't be the trigger that switches. It should be a clear user-specified action. Otherwise, people will complain about the selected trigger-level (i.e. it triggers when zoomed in closer but I want to see the whole US via county-level).

A new UI panel could be created that would allow you to control which grouping to use. It wouldn't include every country with cases. Only those with different groupings.

• China: Province and Country level

• USA: County, State, Country

• Korea: Districts, Province, and Country level (not in this dataset though)

[urbnfntsys1973]

Could change/specify a tolerance of accuracy dependent on zoom level.. Maybe?

[urbnfntsys1973]

Data of this global magnitude is huge, If a County tests positive for a case, then does further testing and determines it is negative (tests are not 99% effective yet), then the margin of error at a smaller level will be greater if data is determined between auto uploads. Not saying this is ideal, but could be why they have chosen to change the scope of the map. Better detail is better but more prone to error.

[ciscorucinski]

@urbnfntsys1973 the dataset still breaks it down via county-level, but the map is using the state-level

[urbnfntsys1973]

@ciscorucinski agreed, but for margin of error on public view may be the reason. I like all here would still prefer the more detailed view. The data is still there you just have to modify their feature script and build your own dashboard using their data to retrieve it. I am looking into that currently. A job better suited for tomorrow.

[jrmatchett]

I've been accessing their data for a few days now to mix with other data and dashboards, and what I've observed is that the feature layer is the dataset that is updated throughout the day, while the individual date and time-series CSV files in this repository appear to be snapshots of the feature layer data taken at 00:00 UTC each day. So we'll see if the next daily snapshots still includes the more granular locations.

[mw32]

@urbnfntsys1973 zoom function is exactly what I was saying above. This is similar to smart chart or those chart libraries often used in js.

[ciscorucinski]

@mw32 @urbnfntsys1973 the problem with a zoom trigger is when you want to view county and state-level data on the same zoom levels. You just cannot do that. Zoom-level triggers work best if you want to focus on one point (I want to rent a home in LA and want to see what's best in that limited view ... you are drilling down to a very specific location). We don't always want that here. We want to compare across the US, across countries. Zoom-level triggers don't help with that.

I think it would be more beneficial for everyone to allow the viewers to decide the groupings. What to see the wide-view, county-level data. Good, you can do that. What do view the wide-view state-level data to compare with Canada? Good, you can also do that. Want to look at just the West Coast and see both the county and state-level comparisons? Good, that is also possible. Plus, then this can be expanded to China and other countries if they get lower-level data.

You get what you want. I get what I want. Everyone gets what they want. That doesn't happen with many other methods.

On a side note, how about this ... there are 23 zoom levels. Zoom all the way out to level 0, and then count how many zooms should show the state levels (zoom 0, 1, 2 ...), and at which zoom level should it switch to county-level?

[EdMabe]

The time_series data still contains latitude and longitude and although it's a bit of work, your can create a custom Google map using the data to get greater granularity. I made this today (please forgive my rudimentary map skills) https://drive.google.com/open?id=10Dw_OZ928gHFoGXWzOZyXR5_QAwXyKZ7&usp=sharing

[jqnatividad]

NYTimes has Hyperlocal data behind its COVID-19 US map https://www.nytimes.com/interactive/2020/world/coronavirus-maps.html#us

Perhaps, they can be convinced to open source it and follow JHU’s lead?

[ciscorucinski]

@jqnatividad It makes complete sense to use county-level reporting with a US map only. Perhaps @CSSEGISandData can create another map. They have accumulated and active maps, they could add a US map

[pnisita]

This is virtually useless without the local data, especially for people in high population large states like NY, Cali, Texas, etc. Here in Long Island, the overall state cases means nothing if we don't easily know how many are on the island itself.

[chucksings]

Like if you request they return to view by county.

[MuffleKerfuffle]

In TX, county-level Data still there in the most recent update. (14 hrs ago) I can't seem to locate where, or which update changed to a compiled state-level result.

[CSSEGISandData]

Please see Issue #382