In v0.4.2 I've changed the standard by which α in estimated, in the basis of subtracting 1 to the calculation of the slopes due to the unjustified assumption that data is always a motion. This naturally lead to negative values of α for noises and any data that is shuffled. Serious faceplant here :|. I've corrected this and now the estimated α should be correct. Note that 'nasty' data might still show negative α -- estimation is a subtle art. This should close issue #34.
In
v0.4.2
I've changed the standard by which α in estimated, in the basis of subtracting1
to the calculation of the slopes due to the unjustified assumption that data is always a motion. This naturally lead to negative values of α for noises and any data that is shuffled. Serious faceplant here :|. I've corrected this and now the estimated α should be correct. Note that 'nasty' data might still show negative α -- estimation is a subtle art. This should close issue #34.I can also close issue my own #2 .