Optimal Proxy Selection for Socioeconomic Status Inference on Twitter

Figure 3

(a) The fraction of time a user spends in his/her top 10 most visited locations per socioeconomic class. (b) , fraction of time spent at the most visited location per socioeconomic class for users with at least different locations. (c) Average radius of gyration per socioeconomic class (error bars are computed as , with number of samples).