Hi SPM Users,
This question has been partly answered before but I am not entirely sure if I am doing this right.
Analysis: I have 2 population groups and I did VBM kind of analysis for CBF data that we acquired. Now, we have spmT_0001.hdr/img.
Problem 1: When we used 3dClustsim in AFNI using NN=3 and smoothing obtained from SPM residuals in SPM.mat file, we found that we need a cluster with at least 50 voxels to have FWE corrected alpha of 0.01 for an uncorrected p of 0.005.
What is the procedure to obtain the clusters using this information preferably using SPM?
I tried xjview and bspmview; loaded my spmT_0001.img and asked the software to provide me with clusters that has an uncorrected p<0.005 but cluster size >=50. Can I use the cluster mask hence generated and write in the paper that the clusters were FWE corrected at alpha of 0.01.
Even MRICRO has the same clusters as generated by the other 2 softwares.
I was hoping that if I pass the voxel coordinates in small volume tab in SPM8 and voxel radius as 50 then SPM will output a table where pFWE-corr should be less than 0.05 but that is not true. What am I missing?
Problem 2: My groups are unbalanced. Since 3dClustSim outputs the cluster for p and not for T (that I have from SPM) I need to go from p to t using my degrees of freedom. Can I simply use df = N1 + N2 - #of regressors or I have to use the df as outlined here : https://en.wikipedia.org/wiki/Student%27s_t-test#Equal_or_unequal_sample_sizes.2C_unequal_variances for unequal sample size population. If Yes, how to I calculate the standard deviations/variances from my SPM maps?
I would appreciate any help regarding my problem.