Abstract
This article presents an analysis of friendship choice data by focusing on homophily (or inbreeding bias) and social distance revealed in the patterns of both association between subjects' and friends' statuses and association among friends' statuses. These two aspects of inbreeding bias and social distance are simultaneously taken into account in modeling the data of friendship choice from subjects with different numbers of friends. The statuses of friends are expressed in terms of their combinations rather than their full cross-classifications. Conditionally symmetric log-bilinear partial association models are usefully employed for the analysis. The structural characteristics of inbreeding bias and social distance are identified by comparing nested models and through the interpretation of parameters estimated from models that adequately fit the data.