Correlating behavior measures with fmri data

Hello, we want to correlate behavior measures with our fmri data.

We’ve used 3dRegAna in the past, but could we also use 3dttest++? Is using 3dttest++ a better way to run the correlation?

If so, how do we set it up to use 3dttest++?

If we include the -setA as the imaging set and -covariates as the behavior measure set, how do we get the output as an r value?

Thanks,

Michael

You can use 3dttest++, but I personally prefer to use 3dTcorr1D.

My (fairly old) notes on using 3dttest++ or 3dMEMA is that you would need to standardize your MRI and behavioral measures beforehand.

3dttest++ is preferred because it’s being updated more frequently, plus it has the recent approach for FWE correction.

If we include the -setA as the imaging set and -covariates as the behavior
measure set, how do we get the output as an r value?

See here: https://sscc.nimh.nih.gov/sscc/gangc/tr.html

Thank you both for the responses.

I have 2 questions:

  1. If I use 3dRegAna or 3dTcorr1D, is it necessary to center my behavior measures? If so, why?

  2. If I use 3dttest++, why do I have to standardize both fmri and behavior measures?

Thanks,

Michael

  1. If I use 3dRegAna or 3dTcorr1D, is it necessary to center my behavior measures? If so, why?

Perform centering only if you’re interested in the intercept in this case.

  1. If I use 3dttest++, why do I have to standardize both fmri and behavior measures?

No, you don’t have to. If you really want to obtain correlation value, check out the link I provided.

That’s really clever to use the t-stat Gang! Makes life considerably easier.

In my (now outdated) way of doing things, z-scoring the data would convert the b-value of the regression to a standardized regression coefficient (beta), which in simple regression is a correlation. So if you centered both of your datasets (MRI and covariates) the correlation value would appear in the “mean” column for the covariate. Whereas if you just convert the t-stat of the covariate from t → R, you get the same value.

Thank you both for your input.

It helped a great deal.

Michael

Dear Gang,
I am trying to do behavioural data correlation with my MRI data, would you please let me know the best way to do this. I tried 3dttest++ as you mentioned here and I get Zscr , can I also run -Clustsim with that? the link you sent before is not working and that would be great if you can help with an example.
BR

I am trying to do behavioural data correlation with my MRI data, would you please let me know the best way to do this.
I tried 3dttest++ as you mentioned here and I get Zscr , can I also run -Clustsim with that?

Yes, it’s fine to use option -Custsim.

the link you sent before is not working and that would be great if you can help with an example.

The link was about conversion from regression coefficient to Pearson correlation coefficient.

Sorry the link isnt working and im unclear on how to get r values from 3dttest++, could you please clarify?

Try this:

https://sscc.nimh.nih.gov/sscc/gangc/tr.html

thank you!