MATLAB for Blockmodel Entropy Significance Test (BESTest)

May 3, 2017 Described in the paper:

"The ground truth about metadata and community detection." Peel, Larremore, Clauset. Science Advances, 2017.

Comments or questions to larremore@santafe.edu

Usage:

BESTest.m should be called as p = BESTest(adjMtx,partition,nSamples,modelName)

OUTPUT

• p - p value as described in the paper

INPUTS

• adjMtx - NxN undirected network adjacency matrix. Code will check to ensure that the matrix is either symmetric or triangular.
• partition - Nx1 or 1xN vector in which partition(i) is an integer that enumerates which group vertex i belongs to.
• nSamples - number of samples used to compute p. Recommended 10k or higher for confident results.
• modelName - string with desired model. Four options:
  • 'SBMpoisson'
  • 'dcSBMpoisson'
  • 'SBMbernoulli'
  • 'dcSBMmultinomial'

Example Code:

see lazegaLawyersDemo.m for sample code and usage, fully reproducing Table 1 from the manuscript