You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This data source is based on the [COVID-19 Search Trends symptoms dataset](https://github.com/google-research/open-covid-19-data/tree/master/data/exports/search_trends_symptoms_dataset). Using this search data, we estimate the volume of searches mapped to symptoms related to COVID-19 such as _anosmia_ (lack of smell) and _ageusia_(lack of taste). The resulting daily dataset for each region shows the relative frequency of searches for each symptom. The signals are measured in arbitrary units that are normalized for population and scaled by the maximum value of the normalized popularity within a
18
-
geographic region across a specific time range. **Thus, values are NOT
19
-
comparable across geographic regions**. Larger numbers represent higher numbers of symptom-related searches.
17
+
This data source is based on the [COVID-19 Search Trends symptoms
18
+
dataset](https://github.com/google-research/open-covid-19-data/tree/master/data/exports/search_trends_symptoms_dataset). Using
19
+
this search data, we estimate the volume of searches mapped to symptoms related
20
+
to COVID-19 such as _anosmia_ (lack of smell) and _ageusia_(lack of taste). The
21
+
resulting daily dataset for each region shows the relative frequency of searches
22
+
for each symptom. The signals are measured in arbitrary units that are
23
+
normalized for population and scaled by the maximum value of the normalized
24
+
popularity within a geographic region across a specific time range. **Thus,
25
+
values are NOT comparable across geographic regions**. Larger numbers represent
The `sum_anosmia_ageusia_raw_search` signals are simply the raw sum of the values of `anosmia_raw_search`
38
-
and `ageusia_raw_search`, but not the union of anosmia and ageusia related searches. This is because the data volume is calculated based on search queries. A single search query can be mapped to more than one symptom. Currently, Google does not provide _intersection/union_ data. Users should be careful when considering such signals.
44
+
The `sum_anosmia_ageusia_raw_search` signals are simply the raw sum of the
45
+
values of `anosmia_raw_search` and `ageusia_raw_search`, but not the union of
46
+
anosmia and ageusia related searches. This is because the data volume is
47
+
calculated based on search queries. A single search query can be mapped to more
48
+
than one symptom. Currently, Google does not provide _intersection/union_
49
+
data. Users should be careful when considering such signals.
39
50
40
51
## Limitation
41
-
When daily volume in a region does not meet quality or privacy thresholds, set by Google, no value
42
-
will be reported. Since Google uses differential privacy, there is artificial noise added to the raw
43
-
datasets to avoid identifying any individual persons without affecting the quality of results.
52
+
When daily volume in a region does not meet quality or privacy thresholds, set
53
+
by Google, no value will be reported. Since Google uses differential privacy,
54
+
there is artificial noise added to the raw datasets to avoid identifying any
55
+
individual persons without affecting the quality of results.
44
56
45
-
The data is normalized by the total number of Search users in certain regions for a certain time period and is scaled considering the maximum value of the normalized
46
-
popularity across the entire published time range for that region over all symptoms. The values
47
-
of symptom popularity are **NOT** comparable across geographic regions. Due to the scaling step,
48
-
most of the values should be in the range 0-1. However, since the scaling factor is calculated and stored at a certain time point, the symptom popularity released after that time point is likely to exceed the previously-observed maximum value which results in values larger than 1.
57
+
The data is normalized by the total number of Search users in certain regions
58
+
for a certain time period and is scaled considering the maximum value of the
59
+
normalized popularity across the entire published time range for that region
60
+
over all symptoms. The values of symptom popularity are **NOT** comparable
61
+
across geographic regions. Due to the scaling step, most of the values should be
62
+
in the range 0-1. However, since the scaling factor is calculated and stored at
63
+
a certain time point, the symptom popularity released after that time point is
64
+
likely to exceed the previously-observed maximum value which results in values
65
+
larger than 1.
49
66
50
67
51
68
## Geographical Aggregation
52
-
The state-level and county-level `raw_search` signals for specific symptoms such as _anosmia_ and _ageusia_ are taken directly from the [COVID-19 Search Trends symptoms dataset](https://github.com/google-research/open-covid-19-data/tree/master/data/exports/search_trends_symptoms_dataset) without changes. We aggregate the county-level data to the MSA and HRR levels using the population-weighted average. For MSAs/HRRs that include counties that have no data provided due to quality or privacy issues for a certain day, we simply assume the values to be 0 during aggregation. The values for MSAs/HRRs with no counties having non-NaN values will not be reported. Thus, the resulting MSA/HRR level data does not fully match the _actual_ MSA/HRR level data (which we are not provided).
69
+
The state-level and county-level `raw_search` signals for specific symptoms such
70
+
as _anosmia_ and _ageusia_ are taken directly from the [COVID-19 Search Trends
without changes. We aggregate the county-level data to the MSA and HRR levels
74
+
using the population-weighted average. For MSAs/HRRs that include counties that
75
+
have no data provided due to quality or privacy issues for a certain day, we
76
+
simply assume the values to be 0 during aggregation. The values for MSAs/HRRs
77
+
with no counties having non-NaN values will not be reported. Thus, the resulting
78
+
MSA/HRR level data does not fully match the _actual_ MSA/HRR level data (which
79
+
we are not provided).
53
80
54
81
55
82
## Lag and Backfill
56
-
Google does not update the search data daily, but has an uncertain update frequency. The delay can range from 1 day to 10 days or even more. We check for updates every day and provide the most up-to-date data.
83
+
Google does not update the search data daily, but has an uncertain update
84
+
frequency. The delay can range from 1 day to 10 days or even more. We check for
85
+
updates every day and provide the most up-to-date data.
provides this data for individual census block groups, using differential
33
+
privacy to protect individual people's data privacy.
34
34
35
35
Delphi creates features of the SafeGraph data at the census block group level,
36
36
then aggregates these features to the county and state levels. The aggregated
@@ -59,10 +59,11 @@ doing so, we make the simplifying assumption that each CBG contributes an iid
59
59
observation to the county-level distribution. `n` also serves as the sample
60
60
size. The same method is used for aggregation to states.
61
61
62
-
SafeGraph's signals measure mobility each day, which causes strong day-of-week effects:
63
-
weekends have substantially different values than weekdays. Users interested in long-term
64
-
trends, rather than mobility on one specific day, may prefer the `7dav` signals since
65
-
averaging over the preceding 7 days removes these day-of-week effects.
62
+
SafeGraph's signals measure mobility each day, which causes strong day-of-week
63
+
effects: weekends have substantially different values than weekdays. Users
64
+
interested in long-term trends, rather than mobility on one specific day, may
65
+
prefer the `7dav` signals since averaging over the preceding 7 days removes
66
+
these day-of-week effects.
66
67
67
68
### Lag
68
69
@@ -77,12 +78,17 @@ additional day for SafeGraph's data to be ingested into the COVIDcast API.
77
78
***Number of data revisions since 23 June 2020:** 0
78
79
***Date of last change:** never
79
80
80
-
Data source based on
81
-
[Weekly Patterns](https://docs.safegraph.com/docs/weekly-patterns) dataset. SafeGraph provides this data for
82
-
different points of interest ([POIs](https://docs.safegraph.com/v4.0/docs#section-core-places)) considering individual census block groups, using differential privacy to protect individual people's data privacy.
individual census block groups, using differential privacy to protect individual
86
+
people's data privacy.
83
87
84
-
Delphi gathers the number of daily visits to POIs of certain types(bars, restaurants, etc.)
85
-
from SafeGraph's Weekly Patterns data at the 5-digit ZipCode level, then aggregates and reports these features to the county, MSA, HRR, and state levels. The aggregated data is freely available through the COVIDcast API.
88
+
Delphi gathers the number of daily visits to POIs of certain types(bars,
89
+
restaurants, etc.) from SafeGraph's Weekly Patterns data at the 5-digit ZipCode
90
+
level, then aggregates and reports these features to the county, MSA, HRR, and
91
+
state levels. The aggregated data is freely available through the COVIDcast API.
86
92
87
93
For precise definitions of the quantities below, consult the [SafeGraph Weekly
|`restaurants_visit_num`| The number of daily visits to restaurant-related POIs in a certain region |
95
101
|`restaurants_visit_prop`| The number of daily visits to restaurant-related POIs in a certain region, per 100,000 population |
96
102
97
-
SafeGraph delivers the number of daily visits to U.S. POIs, the details of which are described in
98
-
the [Places Manual](https://readme.safegraph.com/docs/places-manual#section-placekey) dataset.
99
-
Delphi aggregates the number of visits to certain types of places, such as
100
-
bars (places with [NAICS code = 722410](https://www.census.gov/cgi-bin/sssd/naics/naicsrch?input=722410&search=2017+NAICS+Search&search=2017)) and restaurants (places with [NAICS code = 722511](https://www.census.gov/cgi-bin/sssd/naics/naicsrch)). For example, Adagio Teas is coded as a bar because it serves alcohol, while Napkin Burger is considered to be a full-service restaurant.
101
-
More information on NAICS codes is available from the [US Census Bureau: North American Industry Classification System](https://www.census.gov/eos/www/naics/index.html).
103
+
SafeGraph delivers the number of daily visits to U.S. POIs, the details of which
0 commit comments