3dClusterize: I'd like positive only. 1sided or bisided?

AFNI version info (afni -ver):
Precompiled binary linux_openmp_64: Aug 23 2023 (Version AFNI_23.2.06 'Marcus Didius Severus Julianus')

I'd like to get exclusively positive clusters from 3dClusterize. Would it be better for me to use 1sided testing?

(in addition, if 1sided, should I adjust the p-value? i.e. if 0.05 for bisided, should I do 0.025 for 1sided?)

If you are doing a 1-sided hypothesis, you can just use 3dClusterize -1sided .. without having to adjust. If you want specifically the positive tail, then 3dClusterize -1sided RIGHT_TAIL .. and the direct threshold value or p-value.

If you are using 3dClusterize after 3dClustSim, note that 3dClustSim provides output for 1-sided tests, so you can use that for -clust_nvox ...

But note that if you are really using a 2-sided hypothesis and just want to select out the positive clusters for viewing, do a full 2-sided clusterizing, and then just use 3dcalc on the results to pick out the positive clusters. If you don't, the odds are quite high of having double FPR from the stated value. See this paper for more about that:

  • Chen G, Cox RW, Glen DR, Rajendra JK, Reynolds RC, Taylor PA (2019). A tail of two sides: Artificially doubled false positive rates in neuroimaging due to the sidedness choice with t-tests. Human Brain Mapping 40:1037-1043.