I have a question regarding POI data. I looked at the monthly patterns of the core locations such as Tufts university, Harvard University and MIT. Why are these numbers so low?

I have a question regarding POI data. I looked at the monthly patterns of the core locations such as Tufts university, Harvard University and MIT. Here are daily checkins for Tufts: [362,325,346,247,254,343,344,336,259,246,160,72,219,139,168,237,264,345,787,169,113,146,123,106,42,30,46,119,137,152,136]
The Cambridge Queen’s Head,45 Quincy St Harvard University,Cambridge,MA,02138 - [37,39,25,21,37,31,37,37,44,40,36,35,45,41,38,30,21,24,5,8,0,1,4,7,18,13,11,25,12,31,25] and MIT: “[1,1,0,0,0,0,0,1,2,1,0,1,2,1,0,0,0,0,0,0,1,0,0,1,0,1,1,0,0,0,0]”. Why are these numbers so low?

In general, some POIs in the SafeGraph data have anomalously low visits, due to misclassification, changes in POI information, or something strange with the local cell phone towers.

In specific application to colleges, however, the POIs labeled, say, Harvard, in the data generally do not make up the entirety of the Harvard campus. If you like, you can see my paper here, which describes how I estimated campus-level visits at colleges using SafeGraph: http://nickchk.com/Huntington-Klein_2020_Which_College_Campuses_Opened.pdf

Thank you! I will take a look at it. That is really useful! My problem is that we are using the social distancing dataset for this project: US Mobility Maps — EndCoronavirus.org… And now I am beginning to understand where the social distancind dataset comes from.

You might be able to (elsewhere) find a mappimg between college campuses and the CBGs they occupy, and calculate distancing in those CBGs. It wouldn’t be campus exactly but that would tell you distancing behavior in and immediately around campuses.

You are right, I can map campuses to CBGs and check values in the Social distancing dataset. I think I know what could be the problem with MIT. MIT has its own WIFI. Phones automatically switch to their WIFI. This could cause a problem.

That’s a good question actually - do phones continue to ping cell towers when on wifi? I think they might

I consulted one company that was developing software for restaurants. Their problem was that they could not get mac addresses of the phones when they switched to wifi in restaurants. It could be a similar problem here.

Hi @Olga_Buchel_NECSI, I believe this issue is an artifact of the POI being enclosed and how our algorithm handles that. In the case of college campuses, the algorithm intentionally excludes the indoor POI from receiving visits because the horizontal accuracy of devices deteriorates within large building structures like shopping malls, pharmacies/banks within grocery stores or hospitals; there’s no way to confidently discern which POI a person visited. For better transparency about when these cases have been identified, Safegraph started including the geometry column “enclosed” in June 2020. This shows up a lot with medical facilities where parents/children have the same NAICS code because we could not find a better fit (e.g. a specialized area of care in the wing of a large hospital). Just some of the nuances to consider.

Hi @todd_hendricks but MIT has a set of separate buildings that do not overlap with other businesses. I think after we map the data on the map, we might see a better picture of what is going on.