The most promising class of statistical models for expressing structural properties social networks observed at one moment in time is the exponential random graph (ERGMs), also known as p* models. strong point these that they can represent a variety tendencies, such transitivity, define complicated dependence patterns not easily modeled by more basic probability Recently, Markov chain Monte Carlo (MCMC) algorithms have been developed produce approximate maximum likelihood estimators....

A class of statistical models is proposed for longitudinal network data. The dependent variable the changing (or evolving) relation network, represented by two or more observations a directed graph with fixed set actors. evolution modeled as consequence actors making new choices, withdrawing existing on basis functions, and random components, that try to maximize. Individual dyadic exogenous variables can be used covariates. change in stochastic result effects (reciprocity, transitivity,...

AbstractA statistical approach to a posteriori blockmodeling for digraphs and valued is proposed. The probability model assumes that the vertices of digraph are partitioned into several unobserved (latent) classes distribution relation between two depends only on which they belong. A Bayesian estimator based Gibbs sampling basic not identified, because class labels arbitrary. resulting identifiability problems solved by restricting inference posterior distributions invariant functions...

A recurrent problem in the analysis of behavioral dynamics, given a simultaneously evolving social network, is difficulty separating effects partner selection from influence. Because misattribution to influence, or vice versa, suggests wrong conclusions about mechanisms underlying observed special diligence data advisable. While dependable and validmethod would benefit several research areas, according best our knowledge, it has been lacking extant literature. In this paper, we present...

The concept of explained proportion variance or modeled is reviewed in the situation random effects hierarchical two-level model. It argued that proportional reduction (estimated) components not an attractive parameter to represent joint importance explanatory (independent) variables for modeling dependent variable. preferable instead work with mean squared prediction error predicting individual values (for at level 1) and group averages 2). shown when predictors are added, defined this way...

Multiple regression quadratic assignment procedures (MRQAP) tests are permutation for multiple linear model coefficients data organized in square matrices of relatedness among n objects. Such a structure is typical social network studies, where variables indicate some type relation between given set actors. We present new method (called "double semi-partialing", or DSP) that complements the family extant approaches to MRQAP tests. assess statistical bias (type I error rate) and power five...

A class of models is proposed for longitudinal network data. These are along the lines methodological individualism: actors use heuristics to try achieve their individual goals, subject constraints. The current structure among these continuous time Markov chain that can be implemented as simulation models. They incorporate random change in addition purposeful follows from actors' pursuit and include parameters must estimated observed Statistical methods estimating testing also used parameter...

The hierarchical linear model approach to a two-level design is considered, some variables at the lower level having fixed and others random regression coefficients. An approximation derived covariance matrix of estimators coefficients (for higher level) under assumption that sample sizes either are large enough. This expressed as function parameters occurring in model. If research planner can make reasonable guess these parameters, this be used guide choice level.

Statistical models for social networks as dependent variables must represent the typical network dependencies between tie such reciprocity, homophily, transitivity, etc. This review first treats single (cross-sectionally observed) and then dynamics. For networks, older literature concentrated on conditionally uniform models. Various types of latent space have been developed: discrete, general metric, ultrametric, Euclidean, partially ordered spaces. Exponential random graph were proposed...

Abstract The complex interplay between bullying/victimization and defending was examined using a longitudinal social network approach (stochastic actor-based models). (co)evolution of these relations within three elementary schools (Grades 2–5 at Time 1, ages 8–11, N = 354 children) investigated across time points year. Most bullies defenders were in the same grade as victims, although substantial number other grades (most often one higher). Defenders usually gender whereas most boys, with...

This article discusses the stochastic actor-oriented model for analyzing panel data of networks. The is defined as a continuous-time Markov chain, observed at two or more discrete time moments. It can be regarded generalized linear with large amount missing data. Several estimation methods are discussed. After presenting evolution networks, attention given to coevolution models. These use same approach chain small number points, but now an extended state space. space be, example, combination...

Abstract A multilevel approach is proposed to the study of evolution multiple networks. In this approach, basic process assumed be same, while parameter values may differ between different For network process, stochastic actor-oriented models are used, which parameters estimated by Markov chain Monte Carlo methods. This applied effects delinquent behavior on friendship formation, a question long standing in criminology. The studied empirically 19 school classes. It concluded that there...