Web Survey Bibliography
Social network data are collected in a variety of ways and are then analyzed in terms of structural properties and network processes. Yet collected social network data, most likely, contain different types of errors. One widely used technique for delineating structural patterns of relationships is blockmodeling. We do not know how vulnerable these methods are to missing data problems. Given that they are positional - focusing on actor locations defined for the whole network - it is reasonable to expect that blockmodeling is highly vulnerable. We focus on actor non-response, as one form of missing data, and data processing strategies designed to treat such missing data. We examine the impacts of these data processing strategies on the results of blockmodeling by using simulated and real networks. A set of 'known' networks are used, errors due to actor non-response are introduced and are then treated in different ways. Blockmodels are fitted to these networks and compared. The outcome indicator is the level of the correspondence of known block structures and the corresponding block structures of the treated networks. We use the Adjusted Rand Index and the proportion of incorrect blocks to describe the level of the similarity. The amount and type of non-response, as well the treatments of this form of missing data, all have an impact on the resulting blockmodel structures.
Conference Homepage (abstract) / (full text) / (presentation)