Geography and social relationships are inextricably intertwined. As people spend more time online, data regarding these two dimensions are becoming increasingly precise, allowing building reliable models to describe their interaction. In [11], the study of user-contributed address and association data from Facebook shows that the addition of social information produces improvement in accuracy of predicting physical location. First, friendship as a function of distance and rank is analyzed. It is found that at medium to long-range distances, the probability of friendship is roughly proportional to the inverse of distance. However, at shorter ranges, distance does not influence much. Then the maximum likelihood approach is presented to predict the physical location of a user, given the known location of her friends. This method predicts the physical location of 69.1 percent of the users with 16 or more located friends to within 25 mi, compared to only 57.2 percent using IP-based methods.