Conventional hierarchy inference methods typically infer group-level hierarchies directly from interactions, and thus rely on a relatively even representation of individuals, and dyads, in the data. However, rates of agonistic interactions may vary between dyads in association with dyadic (e.g. kinship, pair bond status) or individual (e.g. age or reproductive status) attributes. Thus, current hierarchy inference methods may result in some individuals appearing to be either more or less dominant at the group level (i.e. hierarchy) than they actually are (i.e., than their dyadic interaction outcomes would reflect)-merely as a consequence of how often they are observed to interact with certain group members-in structured groups. We tested this using three hierarchy inference methods: the first two, randomised Elo ratings and I&SI are popular methods that builds hierarchies directly from interactions, while the third, Percolation and Conductance, is a newer method that initially builds dyadic relationships from interaction data, and then uses these dyadic relationships to infer the hierarchy.
The code provided here allows the simulating of outcomes of interactions in animal groups with sex-stratified dominance hierarchies, where males (the dominant sex) either i) tolerate, and thus interact with, all females equally or ii) tolerate only breeding females, whereby males redirect their intersexual interactions towards non-breeding females and concomitantly causing a reduction in the rate of interactions between breeding and non-breeding females (as might occur if breeding, but not-non-breeding, females associate closely with males, e.g. being tolerated at monopolisable resources by males). Simulation parameters that can be modified, include: number of times the simulation is run, number of males, number of females (note that no. males + no. females cannot currently exceed 26), proportion of breeding females (this should correspond to an integer when combined with the number of females), hierarchy steepness, sex-specific dyadic interaction rates, whether changes in the rank order are compared only within females or across the entire hierarchy (given that females could be over-inferred to be dominant to males) and lastly the hierarchy inference method.
The consequences of male-female interaction rates for breeding and non-breeding females' inferred hierarchy positions are then illustrated and saved alongside the simulation output.