This changes the API so that the summation dimension (usually trials) comes first, then the channels dimension, then other dimensions (time, frequency, etc.). Everything is faster!
Still to do:
[ ] Figure out what to do about multitaper, where this approach doesn't work. This shouldn't be too hard for things that can be efficiently expressed as accumulation. Maybe harder for Coherence. This is why tests currently fail.
[ ] Tests for two nd inputs
[ ] Change input order for spike-field statistics
[ ] Faster jackknife for PLI, JammalamadakaR, and JuppMardiaR
This changes the API so that the summation dimension (usually trials) comes first, then the channels dimension, then other dimensions (time, frequency, etc.). Everything is faster!
Still to do:
PLI,JammalamadakaR, andJuppMardiaR