search

hdt-modeling / hdt-forecast

0 stars 0 forks source link

imputing values for mobility data #10

Closed sxsheng closed 3 years ago

sxsheng commented 3 years ago

Since the SafeGraph mobility data has a 3 day lag time, what are the best ways to impute these 3 days worth of data?

https://cmu-delphi.github.io/delphi-epidata/api/covidcast-signals/safegraph.html

One possible way is to mine data using Google Maps API and use information from that to infer those missing mobility values. Or it can itself be a "local" variable that can be fine tuned.

https://github.com/m-wrzr/populartimes

In terms of coverage it should be fine. The percentage of Americans who own a smartphone is around ~80%, it is presumably higher in a college town. And assuming most individuals aren't cautious enough to disable Google's location services then they're whereabout are constantly logged.

We do need to clarify what the terms of service for Google Map's API are before we proceed. There are alternative ways for scraping the data without using their API.

sxsheng commented 3 years ago

Apple and Google contact tracing app might be another possibility? Coverage may be an issue.