SignedGroupModel

Elizabeth Sander and Stefano Allesina

Overview

This package contains code to search for and analyze group model structure of signed adjacency networks.

Search Algorithm

All code to search for network groupings can be found in the CCode folder. To compile the code, run make from terminal while in the folder. See CCode/example.sh for an explanation of how to run the algorithm.

We use a Metropolis-Coupled Markov Chain Monte Carlo (MCMCMC) search algorithm with a Gibbs sampler to search for the optimal grouping for a given network. Bayes factors are used for model selection.

Collecting Taxonomic Data

Code to create taxonomic records for a species list from the ITIS database is in the PythonCode folder. Detailed instructions can be found in PythonCode/README-GetTaxonomy.txt.

Partition Analysis

Several functions to compare and analyze network groupings can be found in the RCode folder. MutualInformation.R contains functions to calculate the similarity between partitions. JackknifeMIs.R calculates the MI between two partitions relative to the maximum possible MI. RandomizationPvals.R calculates the p-value associated with a MI value (here, the p-value is the probability of getting an equal or higher MI when the group identities in the partitions are randomized).

License: MIT