Our method for generating surrogate graph signals is based on sign-randomization of the graph Fourier coefficients. As such, surrogate graph signals exhibit the same graph correlation structure as the original data. The method of surrogate data can be widely applied for non-parametric statistical hypothesis testing.

[1] E. Pirondini^*, A. Vybornova^*, M. Coscia, and D. Van De Ville, “Spectral Method for Generating Surrogate Graph Signals”, IEEE Signal Processing Letters, vol. 23, no. 9, pp. 1275-1278, 2016.

Matlab code