Using hierarchical joint models to study reproductive interactions in plant communities

Pollinator‐mediated reproductive interactions among co‐flowering plant species are prime examples of how species interactions may affect fitness and community assembly. Despite considerable interest in these issues, statistical methods for assessing signal of reproductive interactions in observational data on co‐flowering species are currently lacking. We propose a flexible method for quantifying potential reproductive interactions among co‐flowering plant species using the hierarchical latent‐variable joint models implemented in the Hierarchical Modelling of Species Communities (HMSC) framework. The method accommodates any measure of reproductive success, including pollinator visitation, stigma pollen loads, and seed set. We demonstrate the method by analysing a dataset on bumblebee visitation to a set of co‐flowering plant species in a species‐rich meadow in Norway, and provide R tutorials for this and additional data types. The example analysis revealed both positive and negative effects of heterospecific flower abundances on visitation to co‐flowering species, which we interpret as potential reproductive interactions. Synthesis . Hierarchical joint models provide a flexible approach to analysing patterns of covariation in the reproductive success of co‐flowering species, thus identifying potential species interactions. Important strengths include explicit consideration of community‐level effects and the assessment of residual fitness correlations after controlling for covariates such as flower abundances and phenotypic traits, yielding more complete insights into pollinator‐mediated reproductive interactions.