Is this API here the same that supports FB visualization? about delphi-epidata

[dfarrow0]

Hey fellow character!

This API surfaces the same data that's shown by that visualization. For geo_type=county, geo_value is a FIPS 6-4 county code, which is distinct from zip code.

See https://www2.census.gov/geo/docs/reference/codes/files/national_county.txt for a list of county codes. Find more info on the census website.

From the file above:

CO,08,031,Denver County,H6

Denver county has FIPS 6-4 code 08031. So you could use the following link to get data in that county:

https://delphi.cmu.edu/epidata/api.php?source=covidcast&data_source=fb-survey&signal=cli&time_type=day&geo_type=county&time_values=20200401-20200420&geo_value=08031

[berniefu]

lol. Thanks David!

To confirm another detail, is the unit of the 'value' field percentage?
"value":1.83148395307
^
|
a.k.a. is this 1.83%?

[dfarrow0]

For this particular data_source=fb-survey, yes, the value is a percentage. In general, the unit depends on the data source.

[berniefu]

Reopen this again to confirm more details: I calculated the county level data from 4/1 to 4/20 (getting a mean from value and sample_size), and generated a tableau graph on my side. However, the data doesn't not quite look at the FB map (https://covid-survey.dataforgood.fb.com/):

Could you please share more details about how the FB data is generated?

Specifically as an example, for FIPS 27137 (St Louis MN), there is only one data point in the time range:
https://delphi.cmu.edu/epidata/api.php?source=covidcast&data_source=fb-survey&signal=cli&time_type=day&geo_type=county&time_values=20200401-20200420&geo_value=27137

The value is 0.145, which is different from the FB data (0.77% at this moment).

Thanks!!

[dfarrow0]

I'm most familiar with the technical aspects of the API. I'm not as comfortable with questions on how this particular signal is generated.

@lcbrooks would you be able to answer this question?

[brookslogan]

Facebook's visualizer's county-level data is based on weekly estimates using the weighted version of the estimator. It should be most similar to the smoothed_wcli & smoothed_wili signals.

[berniefu]

Thanks Brooks! There is a wili signal types for FB survey data:
https://github.com/cmu-delphi/delphi-epidata/blob/master/docs/api/covidcast.md
Is that what you meant? I also didn't find anything talking about the 'smoothed'.

[dfarrow0]

The docs were stale. I just updated them with current signal names, including "smoothed" versions.

Relatedly, the covidcast_meta endpoint will return the current set of signal names for each data source (along with other info).

[berniefu]

Thanks both! This is very helpful. I'm now able to get something that's not very far from the fb dashboard (using smoothed_wcli, last 7 day data):