Discussions

I had a few questions about the new Advan Weekly Patterns Plus data.

(1) I plotted total restaurant visits in Chicago on Tuesday evenings and Friday evenings, and I found that there was a drastic and persistent drop in evening visits at the start of 2023. Was there a change in how the hour each visit began in was recorded starting in 2023? The documentation only says that the local time zone is used, but notably, when I make the same plot, but do total visits on Tuesdays and Fridays (i.e., including all 24 hours in the day), there is no such drop.

For reference, I am defining "evening visits" as the total visits recorded in the "VISITS_BY_EACH_HOUR" variable for the hours 5pm through 11pm -- so for Tuesday, I sum over the indices 41 through 47 (i.e., 24 + 17 through 24 + 23) and for Friday I sum over the indices 113 through 119 (i.e. 4 x 24 + 17 through 4 x 24 + 23). If useful, I can send the figure (just let me know how).

(2) I noticed that in the Weekly Patterns Plus data raw counts are no longer provided. Previous discussions on this board about the old Patterns datasets pointed out that if you want to look at small geographies, you need to make some additional normalizations, because the provided normalization is focused on matching the national population -- but this requires the raw device counts. Does Patterns Plus now address geographical biases in the sample, in the provided scaled counts? Or, how should one appropriately adjust the provided scaled counts to look at, e.g., variation in the total number of visits to restaurants by zip code? I want to ensure that any such differences I observe would be due to actual variation in visits by zip code, and not simply due to some geographical biases in the sample.

(3) To confirm, all previous issues in the Weekly Patterns data relating to shared POIs and the SafeGraph POIs should no longer be a concern in the Weekly Patterns Plus data, correct? (E.g. the issues discussed here: https://docs.deweydata.io/discuss/68d3ff7df5304ddf70c4d270) In particular, is the new methodology to define POIs Advan is using for Weekly Patterns Plus applied consistently back through 2017? I just want to confirm because when I plot total Chicago restaurant visits on Tuesday and Fridays with the Weekly Patterns Plus data, I see a similar pattern as I saw with the Monthly Patterns data, where starting in 2023 visits are very flat. In the Monthly Patterns data, I believed this was likely due to the fact that Advan stopped updating POIs after the SafeGraph surge in POIs. This may just indeed be the pattern, but I want to be certain that this is not driven by any shifting in the POI methodology.

Thank you!