Research Paper: _Assessing the validity of SafeGraph data for visitor monitoring in Yellowstone National Park_

Going to highlight some work from the great people at Penn State University. Big round of applause for these Community members! :clap:

Guangqing Chi

Check out their paper, Assessing the validity of SafeGraph data for visitor monitoring in Yellowstone National Park, here! :point_left:

They assess the validity of using SafeGraph data to analyze visitor patterns to national parks, specifically at Yellowstone National Park. The study demonstrates the value and limitations of SafeGraph’s data. Limitations include biases related to aggregations and international visitors. The main area of added value is in demographic analysis. When combined with existing, more traditional survey/count data to the park, SafeGraph data enabled analysis of temporal and demographic trends. For this use case, SafeGraph data was best suited to be a supplemental dataset.

Very exciting to see this publication! @Yun_Liang_Penn_State_University was very helpful in providing feedback to the SafeGraph team on how we could improve our coverage around national parks. Many thanks to you and and the entire team, Yun!

We’re always looking to improve our data at SafeGraph! If you ever have feedback on any of our datasets, please let us know! We’re always open to adding new brands, POI types, locations, and more to our data.

This topic was automatically generated from Slack. You can find the original thread here.

This is an interesting application of SafeGraph data! I had a few questions:

  1. How are the surveys collected? As far as I am aware, I have never been asked my education level when visiting a national park; are you at all concerned about errors in the survey method, and what are the error bars on the estimates?
  2. Yellowstone, like many national parks, probably has dubious cell service in places. What happens in areas where the cell service cuts out? How does SafeGraph map those non-pings? I imagine that the POIs probably do have cell service, but what happens if you have been hiking, and come back to the Old Faithful General Store?
  3. What were the POIs that didn’t map to Google Maps locations? There were a number of them – nearly as many as did map to Google Maps locations – so I was curious what they were, and what it meant that nearly 50% of POIs were not valid locations.
  4. I recognize this is a descriptive paper, but I was curious to know if you had any intuitions about some of the results you got. For instance, you find that 25-34 year olds are less likely to be represented in the survey data than in the SafeGraph data, which is the opposite of what I would have expected.
    Poor people were also overrepresented in the SafeGraph data, while I would have expected people with income <25k to be less likely to own smartphones than those >150k. Do you have any thoughts on why you got the results you did?