2

Here is my problem:

  • I have two undirected networks, $G1$ and $G2$ which change over time
  • The nodes in each network are identical
  • The edges are always constrained between 0 and 1
  • I want to know whether $G1$ leads $G2$ or $G2$ leads $G1$

What is the best way to approach this?

I have thought about treating this like a vector autoregression on the edge weights, and then conducting tests for granger causality (i.e: Does $G1$ predict $G2$ above lagged values of $G1$)

asd
  • 199
  • 3
  • One could put these networks into adjacency matrices and model time-evolution of those. – msuzen Feb 20 '22 at 14:54
  • Sure, they basically are similarity matrices actually. But how would I go about modeling the time evolution of each as a function of the other? Would it be as I said -- granger causality and vector aggression – asd Feb 20 '22 at 21:52
  • See relevant discussion https://stats.stackexchange.com/questions/563735/inferring-causal-direction/563786 – msuzen Feb 20 '22 at 21:59

0 Answers0