Nestimate: An R Package for dynamic, probabilistic, and high order network analysis

Networks are often estimated from data such as psychometric scales, event data, and ecological momentary assessment data. We introduce the Nestimate R package, which offers a unified interface for estimating networks through a standardized workflow. First, Nestimate estimates psychological network models using correlation, partial-correlation, regularized (graphical lasso), Ising, and mixed graphical networks from both frequency and event data. Second, Nestimate introduces multi-cluster networks, which aggregate psychometric networks using, for example, factor loadings, principal components, or network connectivity. Third, higher-order network models represent dynamics whose probability depends on more than the immediately preceding state and offer a rich interface for measuring and evaluating higher-order behaviors such as critical thinking. Fourth, Nestimate provides functions for estimating dynamic networks, including transition, co-occurrence, and mixed networks, for capturing behavioral dynamics from event data. Every estimated network can be subjected to a unified validation pipeline, including non-parametric bootstrapping, permutation testing, split-half reliability, and case-dropping centrality stability analyses.