Hey Mike
you can see Tom Nichols (co-author on the PNAS paper) comments here
http://blogs.warwick.ac.uk/nichols/entry/bibliometrics_of_cluster/
the bottom line is if you do p=0.05 to find clusters and then use
cluster inference on that, don't expect to have a good control of type 1
FWER - but you knew that, why would you look at data at such weak
threshold in a 1st place .. if you stick to p=0.001 default and then
check the corrected cluster values, you should be fine
cyril
> Dear Guillaume,
>
> Thanks for the article. I am just astonished by their findings about false positive rates because they declared the parametric methods can give a very high degree of false positives (up to 70%, compared with the nominal 5%) for clusterwise inference.
>
> Mike
--
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.
|