statcond 'paired' {'on' 'off'} is not working for mixed designs

[arnodelorme]

I confirmed that statcond should support the option 'paired' {'on' 'off'}to address a mixed design (namely, paired + unpaired), but actually does not seem to compute in that way. At least, the command line use of 'paired' {'on' 'off'} returns error (rejected by finputcheck).

Makoto

[arnodelorme]

Yes, it does not compute it in that way. If one of the dimension is unpaired, it treats everything unpaired.

Doing the bootstrap and permutation one is not too hard. We would need the Matlab function to do this in a parametric way though.

Arno

[arnodelorme]

Shuffling differently in both dimensions of the cell array (one dimension paired and one dimension unpaired) would be almost equivalent to shuffling unpaired in both dimensions. I am not sure it is the correct way to do it and would lead to better stats.