Accounting for missing actors in interaction network inference from abundance data. Issue 5 (28th June 2021)

• Record Type: Journal Article
• Authors: Momal, Raphaëlle; Robin, Stéphane; Ambroise, Christophe
• Is Part Of: Journal of the Royal Statistical Society. Volume 70:Issue 5(2021)
• Page Start: 1230
• Page End: 1258
• Publication Date: 2021-06-28
• Subjects: abundance data -- graphical models -- matrix tree theorem -- missing actor -- network inference -- Poisson log‐normal model -- Variational EM algorithm; Statistics -- Periodicals; 519.5
• DOI: 10.1111/rssc.12509
• Languages: English
• ISSNs: 0035-9254

Abstract

Abstract: Network inference aims at unravelling the dependency structure relating jointly observed variables. Graphical models provide a general framework to distinguish between marginal and conditional dependency. Unobserved variables ( missing actors ) may induce apparent conditional dependencies. In the context of count data, we introduce a mixture of Poisson log‐normal distributions with tree‐shaped graphical models, to recover the dependency structure, including missing actors. We design a variational EM algorithm and assess its performance on synthetic data. We demonstrate the ability of our approach to recover environmental drivers on two ecological data sets. The corresponding R package is available from github.com/Rmomal/nestor .