title          :: AFNI GUI: Clusterizing
presentation   :: 
scripts        ::
date           :: 2020-03
speaker        :: Taylor
outline        :: 
(May be useful to view this after 'View model results' talk!)
Quick review of displaying modeling/statistical results in AFNI GUI
Voxelwise thresholding isn't the end of most results
    + most voxelwise calcs are 'massively univariate analyses' (MUA)
    + need to have additional correction/threshold for so many multiple comparisons
      - do something for overall false positive rate (FPR) reporting
    + in general, 'clusterizing' is popular second level thresholding
      - put more faith in extended regions of voxels that survive voxelwise thr
      - *how* to calculate appropriate cluster size is a different talk (!)
Click 'Clusterize' button in GUI to applying clusterizing parameters
Important practical concerns for finding clusters:
    + what voxels are 'contiguous'?  Sharing a face, or just edge, or node?
      - defined by 'NN' (nearest neighbor) parameter:
        NN = 1 --> voxels share face only
        NN = 2 --> voxels share face or edge only
        NN = 3 --> voxels share face or edge or node
      - no inherent 'right' choice, but be consistent across analysis
      - also, different softwares tend to choose different 'NN' values
      - in AFNI, you can choose...
    + do you want large positive and negative olay values in same cluster?
      - i.e., should positive and negative effects be considered part of 1 region?
      - 'bisided' means positive and negative effects are separated
        ... which probably makes sense in most cases
After clusterizing, hit 'Rpt' button to see report about remaining clusters
    + clusters listed by decreasing size
    + see/jump to a representative location (Peak, Cmass, Icent)
      - but one value/location is not entire story of cluster!
    + 'Flash' button to see a given cluster clearly
    + can load in auxiliary dset (like EPI) and 'Plot' average within each cluster
      - can save ('Write') the average to a 1D text file, as well
    + 'SaveMsk' to save map of clusters as 3D dset
      - also get 3dClusterize command example in terminal
      - can take/use this, save report, save copy of dset vals inside clusters
If you change threshold values, Clusterize info and Rpt (report) info update
Can also apply 'A' (alpha) and 'B' (boxed) functionality to clusterized data