Good morning,

There are some new tools for facilitating quantitative analysis of FATCAT output by using the existing 3dMVM program. In particular, the latter's multivariate modeling capabilities allow one to make use of the network-themed structure of the former.

Thus, one can use something like an entire functional correlation matrix (from 3dNetCorr's *.netcc output) or structural 'connectivity' matrix (from 3dTrackID's *.grid output) in a single group analysis. This is done by: treating the matrix of values like a repeated measures data set; reading in a CSV file of subject descriptions (e.g., dumped from a spreadsheet of ages, test stores, behavioral measures, genetic information, etc.); and combining all that information in a statistical model. For example, one could ask, "Are FA values across the set of WM connections found in the DMN associated with FAS diagnoses, while controlling for age, sex, and education levels?", and other acronym-laden questions. Additionally, follow-up post-hoc tests can be done to check in which of the network's ROIs the statistical relations were most significant (i.e., is there a specific cabal of regions driving the overall relation?).

The main impetus of this functionality is to use the *the whole network's values together* in one statistical model. If you like mathematical statistics, then you will love reading this recent paper describing how the modeling is done:
Chen, Adleman, Saad, Leibenluft, & Cox (2014)
[afni.nimh.nih.gov]
If you don't like mathematical statistics, then you will thank G Chen for writing the 3dMVM program to perform these operations.

The tools for helping to bridge the FATCAT output and 3dMVM input are written in Python, with modest goals of formatting various data together and writing a basic script that can be further modified by the user. A brief description and demo data set with scripts are available. Make sure that you have the newest AFNI build (from Aug. 11), and then you can run:
$ ./@Install_FATMVM_DEMO
to obtain the directory. See the 'FAT_MVM_README.txt' file therein. Hopefully at some point soon, there will also be a paper that applies this analysis to an actual data set publicly available.

These are the first version of these tools (fat_mvm*.py), and hence they can be considered beta-versions. More functionality (and hopefully less dysfunctionality) should occur over time. You can be a part of that process-- questions, comments and other feedback are welcomed.

Cheers,
pt, Gang, et al.