Dear FSL masters,
I am looking for a short term FSL methods consultant who can help us review and respond to suggested revisions of our submitted manuscript: Brain activity unique to orgasm in women: an fMRI analysis. I would be happy to pay for your input.
The statistician brought in by the editor has made the following critiques: If you think you can help, after reviewing these comments below, please email me at [log in to unmask]
Thanks so much!
Nan Wise
(1)"The authors apply a so-called cluster forming threshold to conduct multiple
comparisons correction. Lately, parametric statistical methods based on
cluster inference have been subject to considerable criticism. In
particular, based on simulations and literature reviews, there seem to be
very strong indications that cluster inference creates false positive
results, even at ‘acceptable’ or ‘default’ clustering thresholds (see
Eklund et al. PNAS 2016 and related papers). This, combined with the
already lenient clustering threshold(s), greatly increases the risk of
false positives in the present work.
I therefore urge the authors to check their results for false positives.
The best way to do this would be to use FSL’s non-parametric statistics.
Permutation methods are based on far fewer assumptions and are a more
reliable method to perform cluster inference. Of course, default initial
thresholds must be used to present statistics on cluster size. These new
results should be presented in tables, with coordinates for the centers of
activation."
(2) In regard to our design which compares the first 10 second of orgasm with various epochs including the first ten seconds of stimulation, the mid 10 seconds of stimulation, and then 10 seconds at the beginning and end of the "recovery" period post orgasm-
"Even with the partnered and self-stimulation data sampled together, the
results still rely on 10 samples of 10s fMRI data compared against 10
samples of 10s fMRI data only. This makes it rather dangerous to nullify or
reject previous findings (i.e. the arousal and orgasm related prefrontal
and temporal deactivation found by Georgiadis et al. 2006, 2009, 2010) that
were based on larger samples. In fact, accepting the null hypothesis is
neither possible nor allowed with the ‘classic’ statistics presently used.
It might be added that Georgiadis et al. (Neuroimage 2010) found prefrontal
and medial temporal deactivation with fMRI, which eliminates the argument
that the PET technique because of its poor temporal resolution is somehow
responsible for this phenomenon."
|