Abnormal fluctuation in patterns data

[liu-wx17]

Hello SafeGraph community,

I have two problems about patterns data.

1. I observe a significant increase in home devices recently. As shown in the following figure, number of devices with home location CBG increases by over 50% in 2022 March. Related to this point, I find that the monthly visit to healthcare POIs (about 0.8 million) decreases significantly in this month (after normalized by CBG_population and CBG_devices_residing as suggested in SafeGraph blog). I am wondering if the decrease in visit exhibited is actually due to the abnormal increase in number of devices with home location CBG (as we need to divide visits by number_devices_residing in the normalization).
   

   
   
   
   

2. I find that the weekly pattern fluctuates heavily in 2018 and 2019, as shown below. I am not sure If any other person has reported this and will be grateful for any advice on how to handle this issue or possible reasons for the fluctuation.
   

   

[Jeff_Ho_SafeGraph]

1. The number of devices with a home location has indeed been increasing over the last few months, owing to some fluctations in our panel and the the timescale in which devices “drop out” of our panel over time. As such, using the CBG normalization may be less reasonable for this use case (visits to healthcare POIs). I recommend instead normalizing by total_visits rather than using number_devices_residing. See this resource.

2. The only kind of fluctuation I’ve seen similar to this is for POIs that have a regular periodicity (e.g., weekends being more popular than weekends). I have not seen a variation before the pandemic and aftterward, however. If you provide more details about when the fluctuations occur (e.g., day of week), that may garner some more hypotheses from others.