group instacorr system limits?

AFNI Message Board
1

AFNI version info (afni -ver):

Precompiled binary macos_13_ARM: May  8 2025 (Version AFNI_25.1.08 'Maximinus')

Hi gurus

trying to use GroupInstacorr for the first time. I have a feeling I can get things working just by playing around with different sample sizes but I'm wondering if there are ways to know ahead of time what my system can handle.

Specs on this computer:

Apple M2 Ultra, 64GB memory, 380 GB hard drive space left.

Specs on the data I'm trying to instacorr:

30 subjects of (errts*) resting state, the 3T data was highly accelerated (10+ mins of TR=0.8s, 2mm iso resolution)

I used 3dSetupGroupInCorr which produced a 4.3G .data file. Then I tried to start it up as described in "afni20_instastuff.pdf" and I get this output:

(base) storris@SFC-RW38 groupInstacorr % 3dGroupInCorr -setA thirtyMOTIONsubjs.grpincorr.niml
++ 3dGroupInCorr: AFNI version=AFNI_25.1.08 (May  8 2025) [64-bit]
++ Authored by: Cox, the Mad Correlator
++ GIC: data file thirtyMOTIONsubjs.nii.grpincorr.data found with 4,669,962,570 bytes of data
++ GIC: -setA opened, contains 30 datasets, 164031 time series, 4,669,962,570 bytes
++ GIC: page faulting (reading) data into memory

setA:0123456789.0123456789.0123456789.0123456789.0123456789.!
++ GIC: total .data bytes input = 4,669,962,570 (about 4.7 billion)
++ GIC: --- Be sure to start AFNI with the '-niml' command line option

 ---  [or press the NIML+PO button if you forgot '-niml']

++ Opening NIML socket 'tcp:localhost:53213' to AFNI............................................................................................tcp_connect(connect): Invalid argument
.............................................................................................................................................. ** Connection times out :-(
** 3dGroupInCorr exits: closing connection to AFNI

I'll probably try with 20 subjects now but:

what can i do to get it to work? set up memory allocation differently? set up a longer connection time out?

and what can i do to anticipate this in the future? i thought with 64GB RAM i'd be fine but is there something about Apple M chips maybe?

thanks for your thoughts!

-Sam

2

Was the afni GUI started (with the -niml option) before you ran 3dGroupInCorr? The connection times out after a while, and so it is best to start afni -niml before 3dGIC.

If you did start afni first, then try again using 3dGroupInCorr -VERB which will print out much more voluminous progress messages. The messages near the point of failure might help diagnose the problem. Or perhaps Rick Reynolds will be able to magically figure it out, once the weekend is over.

3

whoops! my bad. i was reading page 22 of the pdf as start 3dGroupInCorr first and then only later do afni -niml. shucks. n=30 works great! thanks again for the stellar tool, Bob

4

You are welcome. It was inspired by the speed of the single-subject InstaCorr in the AFNI GUI, and by Alex Martin's autism study. It is possible to interactively make scatterplots from the GIC results - but I can't remember how exactly now. That was a long time ago now.

However! The @Install_InstaCorr_Demo script will download some sample data and instructions. I think the vol/ subdirectory has some instructions. If not, perhaps the all knowing Google can help.

5

Okay, I looked up the scatterplot of individual subject dataset correlation results vs. a covariate. For example, taking the individual InstaCorr results (computing inside 3dGIC) and plotting their collection vs. subject verbal fluency. The 3dGroupInCorr help output mentions how to do this, with the -sendall option. Inside the AFNI GUI, you use the Clusterize function to make the scatterplot. I recall that I added this feature for Rasmus Birn. And I recall that it was lots of fun. Good luck, Sam!