Premium
The structure of probabilistic networks
Author(s) -
Poisot Timothée,
Cirtwill Alyssa R.,
Cazelles Kévin,
Gravel Dominique,
Fortin MarieJosée,
Stouffer Daniel B.
Publication year - 2016
Publication title -
methods in ecology and evolution
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 3.425
H-Index - 105
ISSN - 2041-210X
DOI - 10.1111/2041-210x.12468
Subject(s) - probabilistic logic , computer science , exploit , representation (politics) , variance (accounting) , theoretical computer science , sampling (signal processing) , realization (probability) , data mining , graph , machine learning , data science , artificial intelligence , mathematics , statistics , computer security , accounting , filter (signal processing) , politics , political science , law , business , computer vision
Summary There is a growing realization among community ecologists that interactions between species vary across space and time and that this variation needs to be quantified. Our current numerical framework to analyse the structure of species interactions, based on graph‐theoretical approaches, usually do not consider the variability of interactions. As this variability has been show to hold valuable ecological information, there is a need to adapt the current measures of network structure so that they can exploit it. We present analytical expressions of key measures of network structured, adapted so that they account for the variability of ecological interactions. We do so by modelling each interaction as a Bernoulli event; using basic calculus allows expressing the expected value, and when mathematically tractable, its variance. When applied to non‐probabilistic data, the measures we present give the same results as their non‐probabilistic formulations, meaning that they can be generally applied. We present three case studies that highlight how these measures can be used, in re‐analysing data that experimentally measured the variability of interactions, to alleviate the computational demands of permutation‐based approaches, and to use the frequency at which interactions are observed over several locations to infer the structure of local networks. We provide a free and open‐source implementation of these measures. We discuss how both sampling and data representation of ecological networks can be adapted to allow the application of a fully probabilistic numerical network approach.