Network data for macroalgal epiphyte-host interactions in the Caribbean: structure, vulnerability, and conservation implications

This dataset comprises the complete analytical results supporting the findings of Jover et al. (2026) on the structure and vulnerability of epiphyte-host networks in Cuban coastal marine ecosystems. It provides a comprehensive resource for studying bipartite ecological interactions, network topology, habitat modeling, and integrated vulnerability assessments in a biodiversity hotspot.The core of the dataset is a bipartite, undirected network comprising 396 nodes (130 host macroalgae and 266 epiphytic species) and 1,089 documented interactions, compiled from a systematic literature review. It includes species and habitat identifiers, presence-absence matrices, and a suite of network-derived metrics.Key data components are:Network Structure: Global topology metrics (density, modularity Q=0.4239, path length) and per-node attributes (degree, betweenness and closeness centrality) calculated with igraph v1.5.1 in R. The 18 ecological modules detected by the Louvain algorithm are also provided.Habitat Modeling: Inferred habitat distributions for all 130 host macroalgae across eight habitat types (e.g., coral reef, seagrass, mangrove), derived from epiphyte co-occurrence data using a threshold model (τ=0.30) and a hierarchical Bayesian model implemented in Stan.Epiphyte Generalism: Habitat breadth and a generalism index for all 321 epiphytic species, based on their presence across the eight habitat types.Statistical and Simulation Outputs: Results from AIC-based degree distribution model comparison (identifying a truncated power-law as the best fit), network robustness simulations (targeted vs. random host removal), and Monte Carlo permutation tests for niche overlap among central vs. peripheral hosts.Vulnerability Analysis: An integrated multi-criteria vulnerability index (V_total) for each of the 18 ecological modules. This index combines the Marine Habitat Composite Vulnerability Index (ICVHM) for each habitat with a network-based host vulnerability index (V_red), calculated from species degree and betweenness centrality. This analysis identifies key "ecosystem engineer" species and conservation priority ranks for all modules.Modeling Code & Diagnostics: The dataset includes full model comparison tables (AIC, WAIC) for epiphytic richness models (Poisson, Negative Binomial GLM, Bayesian Negative Binomial) and MCMC convergence diagnostics (R-hat, n_eff) for all Bayesian models, ensuring analytical transparency.This dataset will be of value to marine ecologists, network scientists, and conservation biologists interested in species interactions, ecosystem functioning, and vulnerability assessments in tropical coastal environments. The associated R analysis code is available in a public GitHub repository: https://github.com/AJoverCapote/EpiphyteHost-Cuba-Analysis. The dataset is released under a CC BY 4.0 license.