Modeling Multivariate Sequential Dyadic Interactions

This study explores two methods for analyzing sociometric data measured on several relations observed at several points in time. The multirelational, sequential data may be represented in a four-dimensional actors X partners X relations X time points super-sociomatrix. One current means of analyzing such data would be the multivariate, sequential model extensions of log-linear models for relational data. These methods are briefly reviewed in this paper. Because super-sociomatrices can be quite large, these methods are not always practical. In this paper, we seek alternative methods for analyzing such complicated data sets that may be more feasible. In particular, we explore two methods. The first method proposed as an alternative to analyzing such super-socio-matrices is an application of a four-mode eigenvector model. The second proposed alternative method is an analysis of variance applied to parameter estimates from simple log-linear network models. These methods are described in detail and then applied to two real data sets: the relations in a monastery (Sampson 1968), and the friendship ties among a set of college students (Newcomb 1963).