Warnings regarding Correlation Matrix

Version AFNI_21.2.04 'Nerva':

Hi, my QC output has the error message shown below, I’ve run multiple subject data with the same afni_proc.py script and have never received this error. Can you please help me figure out what it is?

Warnings regarding Correlation Matrix: X.xmat.1D

severity correlation cosine regressor pair


IDENTICAL: -1.000 0.000 ( 0 vs. 4) Run#1Pol#0 vs. Run#2Pol#0

Thank you very much for your help.

-SM

Hi SM,

That perfect anti-correlation is an entertaining/educational aspect of having exactly 2 runs. The constant polort terms are exactly predictive of each other: when one is 0, the other is 1, making their correlation -1. For 3 runs, that correlation drops to -0.5.

So this isn't a problem at all.

  • rick

Hi Rick,

Thank you for taking the time to answer my question. Just to follow up, this is the first time I've ever received this message, and none of the other participants I've run with the same task (all with two runs) before or after this person has received this warning. Thoughts on this?

Thanks

-SM

I was a little surprised by this actually, as I thought polort terms were not included in the output since 2013. But there was a window in 2021 were I might have put that back in, ending in Oct 2021. So your binaries are from that window. After 10/21, that message should no longer appear (probably so that it would not show something bad but unimportant in the APQC HTML report).

As to why it is not all of the subjects, I suspect you are censoring, but this subject either had no censored time points, or had the same number censored from both runs. Otherwise the correlation would not be quite exactly -1.

Does that seem to be the case?

  • rick

Hi Rick,

It's good to know that. In our case, the censoring was different for each run.

-SM