Hello. I have a few questions regarding the cell phone data that Advan uses.
Clearly there is a difference between the Advan Patterns and legacy Safegraph Patterns cell phone panels, but are they independent samples? If they are not independent samples how much overlap is there?
Is the cell phone data that Advan uses randomly sampled?
Why does there a sudden drop in devices seen starting in January 2023? I know this question has been asked in another thread, but I am still waiting for an answer.
The legacy SafeGraph and current Advan datasets are derived from different panels. It’s impossible to know how much overlap there was between those panels since each data company sources from dozens of providers, but it is likely that there is some overlap. We’ve always encouraged researchers to just use one partner vs. trying incorporate or benchmark both.
Like most panels, the Advan data is a just a sample of the population. Panels usually aren’t completely random since they don’t have access the the full population, so there could always be some inherent bias.
There are some previous notebooks that were created for measuring sampling bias that can likely be applied to the Advan data as well. Google Colab
I can’t speak for Advan, but from what I’ve heard there were a few major mobility sources that went offline at the end of 2022. Mobility providers are always constantly juggling changes in their supply. Based on another thread, I know that Advan is working on a restatement of the home panel files for 2023.