The InstaCorr menu on the Define Overlay control panel has a new entry: 3dTstat.
This menu item will let you interactively run 3dTstat to compute a voxelwise statistic from a time series dataset. There are 43 statistics available.
One new such statistics is MSSD = Mean Square Successive Differences = a measurement of how much the dataset fluctuates between neighboring time points. Related to that is MSSDsqrt, which is just the square root of MSSD, and MASDx, which is the Median Absolute value of Successive Differences (multiplied by 1.4826 to make it comparable to MSSDsqrt).
These changes are in the source code (on github) now, and should be compiled into AFNI binaries over this weekend.
“Quick” follow-up question re: the implementation of MSSD in 3dTstat and the AFNI GUI-- Are there any recommendations about transforming data before trying to calculate MSSD?
The reason I ask is that there are two nice recent papers from groups in (sunny) Miami that use MSSD, where the investigators z-score their data before calculating MSSD. (Paper 1- Moment-to-Moment BOLD Signal Variability Reflects Regional Changes in Neural Flexibility across the Lifespan
and Paper 2- Resting-State Brain Signal Variability in Prefrontal Cortex Is Associated With ADHD Symptom Severity in Children ). This likely could be done post MSSD calculation, but I wondered if folks had worked with the new options and what they were doing with data transformations?
I’m VERY excited this got added to the AFNI universe (the AFNI-plex?)!
I don’t have any such recommendations available. I did the command line addition to 3dTstat in an hour or so, after seeing the idea on Twitter. I added the GUI part a couple days later in an attempt to avoid doing something more important and less pleasant. But I don’t have any experience in using MSSD or similar quantities for analyzing/interpreting brain data time series.