Collecting a large number of alters in egocentric network research: A comparative analysis of three approaches

Abstract

This article presents an analysis of the impact of the number of alters elicited in an ego network on the structural properties of those networks. There continues to be debate about the pros and cons of eliciting a fixed number of alters for each respondent versus allowing the respondent to list as many or few alters as they would like. This article explores a random assignment of respondents to three treatment groups – (1) a fixed number of alters set at 30, (2) a variable number of alters up to 45, and (3) a variable number of alters up to 45 with a 20 alter minimum. The results indicate that, from a non-structural perspective, all levels of emotional proximity, interaction contexts, genders, and ages are consistently sampled across the three treatment groups. At the structural level, the behavior of individual metrics is also largely similar. However, the most significant differences arise in the collective behavior of structural metrics—specifically, in their correlation structure, the amount of redundant information each variable provides, and the diversity and interpretability of the observed structural variability. When a data collection strategy constrains network size, it reduces the sparsity of the correlation matrix, effectively decreasing the number of independent global variables needed to describe network structure and making these global variables less interpretable. In other words, networks constructed with a survey that limits size tend to be more similar to each other, exhibiting less structural diversity and yielding differences that are harder to interpret. However, we discuss how these differences may simply be mathematical artifacts, without necessarily implying a clear advantage in choosing one treatment over another. Finally, we argue that the field needs a targeted study to answer whether the differing numbers of alters listed is a function of network size.