Main Article Content
This article reviews the statistical formulation and substantive applications of multilevel models for personal network data. A personal network is a social network sampled around a focal individual, the ego. Network nodes are the ego and his or her social contacts, the alters; network edges are ties between ego and alters, and ties among alters as reported by ego, usually indicating acquaintance or various forms of interaction. Personal network datasets exhibit a classical multilevel structure, with alters or ego-alter ties (level 1) hierarchically nested within egos or ego-networks (level 2). Hierarchical linear and generalized linear models have been used in the social and the health sciences to analyze these data and explain the variation of outcomes observed on alters or ego-alter ties. The paper presents these models and the assumptions and hypotheses they imply; outlines their main research applications; and illustrates their use by analyzing real-world data on personal networks and social support among Sri Lankan immigrants in Milan, Italy.