With incorrectly large smoothing window: ~80% of neurons show to be significant when using 100ms timebins and ~60% show to be significant when using 5ms timebins
Refer to bootstrap_test.ipynb
With correct smoothing window size (timebin=100, smoothing_window=250, ignore_freq=0.5), 10,000 sample bootstrap, confidence interval of 0.999, ~63% of neurons show to be significant. This is better than the previous 80%, and might not be a bad method going forward.