[FieldTrip] problem with comparecorr
Gabriel Gonzalez Escamilla
ggonesc at upo.es
Fri Nov 15 11:26:52 CET 2013
Dear FildTrip experts,
We have two groups and we are performing within group correlations with ft_sourcestatistics using cfg.method='montecarlo';cfg.statistic='corrcoef'; cfg.measure='stat'; cfg.correctm='cluster'; cfg.clusterstatistic='maxsum'; everything looks good on the analysis of every group. We've got negative correlation in GP1, and no correlation in GP2.
Then we are performing between-group correlations to compare the regression lines by using cfg.method='montecarlo'; cfg.statistic='comparecorr'; cfg.measure='stat'; cfg.correctm='cluster'; cfg.clusterstatistic='maxsum' on ft_sourcestatistics. for which we are getting a difference between the correlations of the groups (GP1 Vs GP2, as input to ft_sourcestatistics a grandavg matrix containing firstly the subjects on GP1 followed by the subjects of the GP2).
This seems to be ok when we perform the between correlation of GP1 Vs GP2, but then we invert the matrix to perform GP2 Vs GP1 (we have checked that the second matrix has the subjects in the correct order, all subjects of the GP2 now followed by subjects of GP1), which should result in the same results as GP1 Vs GP2 but with opposite direction; but instead of showing this, it shows no results.
We have perform 1000 and 2000 montecarlo (MC)-permutations to ensure that this results are not biased by little MC-permutations, but every time we get, that in the first direction we get results, but in the inverse direction we never get any results. Is this somehow normal?
Can anyone give us so guidence on this?
Many thanks in advanced,
Gabriel.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20131115/86bdf3c6/attachment.html>
More information about the fieldtrip
mailing list