Hi there! I was wondering if there are any counties in the US that do not have any POI representation? Any information about geographic coverage for the Places or Social Distancing Metric would be really useful.
Hi @Taylor_Anderson_George_Mason_University, I don’t know of any US counties that are not represented in SafeGraph’s datasets. If you wanted to check, you could compare the first five characters of the poi_cbg field in Monthly Patterns to the county FIPS codes from something like this website. Although I expect coverage to be very, very high.
What kind of questions are you trying to answer with your work? If I know more about your goals, I may be able to provide some better guidance 
@Ryan_Kruse_MN_State Hi Ryan, I am trying to write a short discussion on the geographic completeness of the SafeGraph data (that we use in our analysis) and whether any counties are excluded because their sample size of users is too small or there are no POIs.
@Taylor_Anderson_George_Mason_University Okay, great. I would suggest checking out the Patterns section of the Places Manual. There is a subsection under raw_visit_patterns on geographic bias, along with a link to a Colab notebook that goes into more detail. Additionally, the visitor_home_cbgs section might have some valuable info.
Generally, any exclusions will happen at the CBG level, not the county or POI level. In some remote or sparsely populated CBGs, there may not be many devices in SafeGraph’s panel. For example, in February 2020’s home panel summary, there are 179 CBGs (out of ~220,000) with just one device in SafeGraph’s panel residing there. A lot of the time, CBGs with low numbers of SG devices also have a low population, but occasionally that is not the case.
Perhaps also relevant is the method used to determine home location. If home location cannot be estimated with a high degree of confidence, then no home location is assigned, which means those devices will not show up in SG’s home panel summary. Those devices will still contribute to raw_visit_counts and raw_visitor_counts though.
Is that helpful? Let me know if you have any questions!