Hi all,
I'm running neural network simulations and using granger causality to compare the Granger drive from one population to another. Generally speaking I'm using a code to find the Akaike Info Criteria, and then using that as the lags (order) for GC. The issue is that I'm calculating GC on multiple "identical" simulations (as in, same input parameters, but different outputs), so the AIC is slightly different for each simulation. Specifically, AIC ranging from 30 - 50. The goal at the end is to average the GC curves into one final curve for "experimental" and "control" groups of simulations.
The question I have is whether it's appropriate to use the exact AIC for *each* simulation, or use the average AIC for all of the simulations I'd like to group together. So, for example, if I have 3 simulations with AIC = 30, 40, 50, should I be using 30, 40, and 50, or just use the avg (40) for all 3?
- Badges
- Science topic
- Similar topics
- Biological Science
- Kinesiology
- Running