add spike-equivalence test

[rgutzen]

Quantifying to what degree the exact times of the individual spikes in two datasets differ. This could be used to compare spike sortings of the same raw data or could serve as the initial test for any comparison to exclude the trivial scenario of identical data sets.

In comparison to a reference recording, each spike can be evaluated as

• False Negative (spike is missing)
• True Positive (spike exists and is attributed to the same unit)
• Misclassified (spike exists but is attributed to another unit [may require channel information])
• False Positive (extra spike) [= False Negative if there is no reference and the two samples are equally legit]

http://www.scholarpedia.org/article/Measures_of_spike_train_synchrony

Is it necessary that both datasets have the same number of units?

[rgutzen]

Hierarchy of tests: same number of units; same number of spikes; same timing of spikes

[rgerkin]

Once you get to single neuron (unit) measures, you could consider using the tests (or subclasses thereof) in neuronunit here. We have actually not written tests that look for equivalence of spike trains, preferring to look for equivalence of spike train statistics, but if you look at ISITest, for example, you'll see the basic logic of getting the spike train and doing computations on it.