Statistically valid null distribution tests

[hfgem]

For deviation statistics add a realistic distribution test:

• Use the true data spike trains to calculate neuron-to-neuron correlations in firing rates
• Use the covariance of the neuron activity to generate spike trains using a correlated Poisson process
• Re-run the deviation analyses on this dataset and compare to true data