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.
Data and other code: http://danlarremore.com/metadata
Comments or questions to email@example.com
BESTest.m should be called as p = BESTest(adjMtx,partition,nSamples,modelName)
- p - p value as described in the paper
- 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:
see lazegaLawyersDemo.m for sample code and usage, fully reproducing Table 1 from the manuscript