Doctoral thesis

Stochastic actor oriented model with random effects : simulation based estimation, model evaluation, and implementation of dynamic sets and multisets

  • 2023

PhD: Università della Svizzera italiana

English This thesis advances the field of social network analysis by generalizing the stochastic actor oriented model (SAOM) to allow for the inclusion of random effects, so that the heterogeneity of the individuals can be modelled more accurately. The method of moments estimation, with the model evaluation procedure that are commonly used in the SAOM, are generalized to being able to estimate the parameters of the model when there are random effects. A social network of workers in a tailor shop has been analysed using the SAOM with random effects. The focus of the analysis has been the comparison of different models, with or without random effects, and the difference in the interpretation of the parameters. The algorithm developed for the SAOM has been studied in more detail in a regression set up, also when the parameters are estimated with generalized method of moments. The focus of the research has been on the comparison between different methods to increase the power of test statistics, when the statistics used to estimate the parameters are correlated. Algorithms that are used in social network analysis are often based on simulating the underline network process, that is represented by a discrete dynamic data structure. An efficient R implementation of sets and multisets, based on hash tables, is discussed and applied to network processes whose state is represented by a set, and whose sufficient statistics are stored in a multiset.
  • English
Computer science and technology
License undefined
Open access status
Persistent URL

Document views: 14 File downloads:
  • 2023INF014.pdf: 1