 |
 |
Agree with Michael. You use cluster inference by saying a cluster larger than XXX voxels is statistically significant (corrected for multiple comparison by FWE-controlled, or in other words, clusters that have cluster p FWE < 0.05). Primary threshold determines which voxels you are seeing when you check for cluster sizes and statistics. On Fri, Dec 8, 2017 at 10:23 AM, Michael Spilka <[log in to unmask]> wrote: > P.S. This article might also be helpful: Woo, C. W., Krishnan, A., & > Wager, T. D. (2014). Cluster-extent based thresholding in fMRI analyses: > pitfalls and recommendations. *Neuroimage*, *91*, 412-419. > > On Thu, Dec 7, 2017 at 8:23 AM, Matt T. Richins <[log in to unmask]> > wrote: > >> Dear collective wisdom, >> >> I was given the following request for an imaging paper and I need your >> advice: >> >> "-For your cluster-based correction for multiple comparisons, an >> uncorrected P value of .001 is now recommended based on Eklund et al.'s >> work." >> >> In our analysis, we used a fairly standard threshold of .05 FWE >> corrected. Is using an uncorrected threshold really advisable now? I've >> read the Eklund paper but I still don't understand how ignoring the >> multiple comparisons issue is ok?? Or am I missing something? >> >> Thanks in advance. >> > >