Steps and considerations before using 3dClustsim

AFNI Message Board
1

Dear AFNI experts,

I am trying to use 3dClustsim on a 3dLME or 3dMVM output to do multiple comparison corrections. I know that first I have to use 3dDeconvolve to get the error time series (errts) from each participant. Then, I should use 3dFWHMx with the -acf option to estimate data smoothness and use those parameters in 3dClustsim.

I have some questions about this process:
Firstly, errts is an output from a regression performed by 3dDeconvolve. So, what can I do if I did that regression using 3dDentred? Can I get the errts in an isolated step? if so, should I input each subject 3D+time image before any detrend with the same regressors than in 3dDentred? Is it possible to get errts file without using regressors? I also would like to understand why is necessary to use errts instead the regressed 3D+time image to get acf parameters.

Secondly, when I use 3dFWHMx I get four parameters which appear in the terminal. Are these parameters the average across all volumes/subbriks? Should I put the first three in acf option when using 3dClustsim then? My last question about this topic is, if I get every subject acf parameters, should I put the mean of all these inter-subject acf parameters with the tstat map as an input (using 3dClustsim) ?

If you could help me with these queries, I would greatly appreciate it.

Thanks a lot,

Karel

2

Ps. I forgot to mention that I if I am using correlation map for the group level analysis, Is it possible to use the option -resid in 3dLME to get the residuals for each subject’s correlation map and then calculate the acf parameters from these residuals (coming from subject correlation maps)? or is getting the errts file mandatory? Finally, from which type of data should extract the errts? from subject EPI or subject statistical maps?

Thanks a lot,
Karel

3

what can I do if I did that regression using 3dDentred? Can I get the errts in an isolated step?

3dDetrend is all you did at the individual subject level? If so, you may use the output from 3dDetrend as an approximation.

Are these parameters the average across all volumes/subbriks?

Yes.

Should I put the first three in acf option when using 3dClustsim then?

Yes.

if I get every subject acf parameters, should I put the mean of all these inter-subject acf parameters with the tstat map as
an input (using 3dClustsim) ?

You can use the average ACF parameters across subjects as input for 3dClustSim with a brain mask that has the same resolution as your group analysis results.

if I am using correlation map for the group level analysis, Is it possible to use the option -resid in 3dLME to get the residuals
for each subject’s correlation map and then calculate the acf parameters from these residuals (coming from subject correlation
maps)? or is getting the errts file mandatory?

Either is fine as an approximation for the ACF parameters.

from which type of data should extract the errts? from subject EPI or subject statistical maps?

Probably preferable to use the residual time series.

4

Hi Gang,

3dDetrend is all you did at the individual subject level? If so, you may use the output from 3dDetrend as an approximation.

Yes, it was at the individual subject level. Then, I will use 3dDetrend output as an approximation.

Thank you so much! all the answers helped me a lot.

Best regards,
Karel