Consistency and asymptotic normality of stochastic block models estimators from sampled data (Q2209826)

This paper is a research in the domain of statistical analyses of networks with missing data. The authors start with the presentation of the stochastic block model, SBM, and missing data for SBM, followed by some sampling design examples, definition of the complete-observed log-likelihood, introduction of parametric models and a set of assumptions on the parameter space. The concepts of identifiability of SBM, permutation, equivalence and parameter symmetry as well as distance, set of local assignments, c-regular assignments, class distinctness and confusion matrix are introduced. The local asymptotic normality of the complete-observed model is proved. The main result states that the observed-likelihood ratio behaves like the complete likelihood ratio, up to a bounded multiplicative factor. As a consequence of the main result, the asymptotic behavior of the maximum likelihood estimator, MLE, and the variational estimator, VE, for the incomplete data models are investigated. Appendix A contains proofs for technical results and Appendix B proofs related to the main results. Appendix C contains results related to sub-exponential random variables and Appendix D results related to likelihood ration of assignments.