 |
 |
Hi Jiang, If you'd like a test level of 0.05, use -log10(0.05) = 1.3010. Stricter thresholds are always a good idea, though (this is in general, not only for PALM). If you used -corrmod after entering all the 8 networks, and -corrcon after entering both contrasts, there is no need for further correction. Use the files "mcfwep". These are corrected for all the 8 modalities and 2 contrasts (16 overall). All the best, Anderson On 28 September 2016 at 16:12, Jian Zhang <[log in to unmask]> wrote: > Dear FSL Experts, > > If I set both -corrmod and -corrcon options of PALM for my analysis, > correcting the results across modalities and contrasts, and use -logp > option, saving the output p-values as -log10(p), which value for threshold > should I choose? For example, if I have 2 groups (patients and normal > controls), 8 resting state networks (modalities) for analysis, testing two > tails differences (patients > normal controls and patients < normal > controls), since I already use PALM to correct for p values across > modalities and contrasts, should I just use -log(0.05) = 1.3 as the > threshold number for significant clusters? Or it's still necessary to use > the number of -log(0.05/8*2) = 2.5? > > Thank you very much. > > Jian >