Hi! I am wondering how SafeGraph collects the cellphone data. We want to study if cell phone penetration is lower in low-income and minority dominant areas than in more affluent areas. Would device count for each CBG be a good indicator of the number of cell phones owned in that CBG?
I think that ought to work. SafeGraph determines the home census block group by analyzing at least 6 weeks of data for a cellphone (between the hours of 6pm - 7am) . There is an edge case to be aware of: where the device counts are between 2 - 4 in a CBG (below 2 go unreported, and between 2 - 4 are reported as 4). More details here: Places | SafeGraph Docs
Also if the concern is about sampling bias, take a look this COLAB notebook shared by Ryan Squires from SafeGraph, it works through bias figures per state (as well as at a county level and income based) and you can follow the same process to calculate bias/abs_bias for the current dataset 
@Amartya_Banerjee_Northwestern_University Thank you very much! I believe the SafeGraph data is reliable. My goal is actually to study whether device count for each CBG is representative for the total number of phones in that CBG. For example, if we found less devices in most low-income and minority dominant CBGs than affluent CBGs, can we say cell phone penetration is lower in low-income areas? Is that because other factors, like different population groups may prefer to use different apps which may not be collected by SafeGraph. @Ryan_Fox_Squire_SafeGraph Can you help to answer this question?