We provide MATLAB codes of key functions to perform the phase diagram analysis on muli-variate time-series data, which was proposed in Ezaki et al. Communications Biology 3: 52 (2020).
Since it is difficult to create a pipeline that performs phase diagram analysis automatically for arbitrary data, we provide the key functions and the codes for example analyses using them.
-
example_01_EstimatePMEM.m This example demonstrates how to estimate h and J in the pairwise maximum entropy model (PMEM) from binarized multivariate time-series data (see Eqs. (1) and (2) in Ezaki et al. (2020)). This example also checks if the means and correlations of variables are similar between the empirical data and estimated PMEM.
-
example_02_PhaseDiagram.m This example demonstrates how to sample data from an Ising model (which was generated by the affine-transformation of
$J$ estimated from the input data; Eq. (3) in Ezaki et al. (2020)) and compute physical quantities to draw phase diagrams (see Figs. 1 a-d in Ezaki et al. (2020)). -
example_03_EstimateParticipantMuSigma.m This example demonstrates how to estimate the \mu and \sigma values for each participant using the phase diagrams (see Fig. 2 in Ezaki et al. (2020)).
You can open and test the code on MATLAB Online (MathWorks account required):
We acknowledge Michio Inoue for his valuable suggestions on procedures used in example_03_EstimateParticipantMuSigma.m. We also thank Douglas M. Schwarz for providing an open source function ("intersection.m", Copyright (c) 2017, Douglas M. Schwarz) used and redistributed in this repository in accordance with its LICENCE TERMS.