¥Overview of Statistical Testing of Group Datasets with
AFNI programs
HParametric
Tests:
åAssume
data are normally distributed (Gaussian)
í3dttest (paired, unpaired)
í3dANOVA (or 3dANOVA2 or 3dANOVA3)
í3dRegAna (regression, unbalanced
ANOVA, ANCOVA)
í3dANOVA4 = Matlab script for one-,
two-, three- and four-way ANOVA
HNon-parametric
analyses:
åNo
assumption of normality
åTends
to be less sensitive to outliers (more robust)
í3dWilcoxon (~t-test
paired)
í3dMannWhitney (~t-test
unpaired)
í3dKruskalWallis (~3dANOVA)
í3dFriedman (~3dANOVA2)
íPermutation
test
åLess
sensitive and less flexible than parametric tests
åIn
practice, seems to make little difference
íProbably
because number of datasets and subjects is usually small (hard to tell if data is non-Gaussian when only have a few sample
points)