I am using 3dTproject to do nuisance regression. Sometimes we combine different fixed of regressors (movement parameters, physiological recordings, etc.) and also a band pass filter.
The problem appears when I try to band pass at “0.002 - 0.01” or “0.012 - 0.028” Hz combined with the mentioned fixed regressors.
It seems that band passing at those frequency ranges use at least more than 90% of the D.O.F from the data set. Then, if I have more regressors (movement parameters, physiological recordings, etc.) it is not possible to do the regression because the number of regressors exceeds the number of time points.
Error message:
+ Block #0: 250 time points -- 243 stopband regressors
++ 1 Blocks * 5 polynomials -- 5 polort regressors
+ -- 14 other fixed ort regressors
[7m** ERROR:[0m total number of fixed regressors (262) is too many for 250 retained time points!
How could I deal with this problem if I want to keep using 3dTproject?
Also, in a hypothetical case, If I am interested to compare results between two samples with different regression models: i) The first one was done using fixed regressors (movement, physiological recording, white matter, etc.) ii) and the second was done only with a band pass filter. .
Are these results comparable considering that the conditions in the regressions were different?
If you could help me with these queries, I would greatly appreciate it.
I could swear I replied to your earlier thread about this, but maybe I never actually completed it and made it aaaaaaaall the way down to the troublesome “post” button. Sorry for letting this linger.
Ignoring the general troubles with band passing, it looks like there is simply a decimal place problem in what you are filtering. It seems that you might actually want to fileter at “0.002 - 0.01” and “0.12 - 0.28” Hz, rather than at “0.002 - 0.01” or “0.012 - 0.028” Hz.
On the flip side, the given numbers should not cost so many degrees of freedom, so maybe these numbers are not correct to begin with (e.g. maybe it was 0.012-0.28). Would you mind posting the actual commands used?
Thank you so much for your reply. I had been talking with Bob about this topic, and the problem was more specific than I thought. Even though the problem is solved, I would like to post what happened because this could be useful for other users.
Basically, I was trying to use “sub bands” which are bands with very short frequency ranges. The cause of the error in the program was because I tried to band pass at these very low frequency (0.002 - 0.01 Hz). Unfortunately, that removed all the independent components in the signal, leaving nothing behind. At that point it is not possible to get any valid statistics. So, the given error is convenient in the way to warn the users that their results are not robust at that point, which is something great.
Kind regards,
Karel
The
National Institute of Mental Health (NIMH) is part of the National Institutes of
Health (NIH), a component of the U.S. Department of Health and Human
Services.