extreme different groups

Hi all, 

I'm trying to compare different groups, one group is older than the other one (experimental group mean age 51 yr and control group 33 yr) but the control group has always a higher activation. I suppose it could be related with the age,  isn't it?

I've tried different processing steps, also I've tried introducing age as covariate of no interest, but nothing.

Any idea?

Thank you in advance!




You mean task activation? If the difference regions looks reasonable, the results should be reasonable.

ZANG Yu-Feng



It's a resting state without task. 

But it's so weird. I'm trying to replicate some papers, but I can`t due to the higher activation of the younger group.

I don't know what to do. 

Would you please give some details of your data processing? The term “activation” is more often for task fMRI. For resting-state fMRI, terms “functional connectivity”, “amplitude of low frequency fluctuation (ALFF)”, “regional homogeneity” (measuring the local synchronization) and so on are often used.




Dear Zang, 

Thank you again for your quick answer. You are right, I made a mistake using the wrong. Sorry. 

Pre and post process

G.E. 3T

RT: 2.5 

248 slices

Functional dimensions   64x64x33 vox size 3.75x3.75x4

Structural dimensions 256x156x248 vox size 1x1x0.75

and data steps processing dparsfa -> http://s16.postimg.org/h1yjssks5/dparsf_A.png


Analysis process, based on two-sample t-test between groups for each roi with zroi data. 
(I performed 6 different image roi regions for the FC, as spheres of 6mm from the Insular Cortex)

As showed in several papers, I would find some functional conn between different areas for patients with chronic pain, but instead of this, I'm only able to obtain FC for the control group in such different areas. 

It could be the different age matched groups? I tried controling age as covariate of no interest, but even with this, I obtain similar results. 

So, any suggestion is welcome!
Thank you in advance!

edit: on the upload image you will see only 2 subjects, but I'm really using  39 (20 & 19).

1. Do you mean you have two groups, patients with chronic pain and controls, but the age is not matched, and even differs a lot? How much?

2.It seems you did not regress out the global mean timecourse? You can try that.




Hi again!


1. Right, differs in the meaning that you can't see any significative FC in pain areas in chronic pain subjects compared to healthy controls, usin an uncorrected p of .001   .... that's so weird. 


2. Now I'm performing again the processing steps with global signal. 


It's a good idea in case of data with artifacts to scrub the data? What do you recommend? 

thank you again!


P < 0.001 is rather stringent.


Although you have strong hypothesis on where the ROIs are, reproducibility of results across studies are not so good. One possible reason is that the location, size, and shape of seed-ROI vary a lot from study to study. ROI pair-wise seems to be more challenging than voxel-wise functional connectivity analysis.


For pain study, you may try to analyze the amplitude of low frequency fluctuation (ALFF) by using repeated two-way ANOVA (frequency*group). Hopefully, you would get decreased ALFF in lower frequency band and increased ALFF in higher frequency band in the anterior insular cortex as well as dorsal ACC (something like Malinen et al., PNAS, 2010; Otti et al., 2013, BMC Psychiatry). Please note that the ANOVA was not mentioned in these papers. I will be happy to send you my presentation and also be happy to have a voice talk about it with you.


Yufeng (zangyf@gmail.com)



Thank you so much!

It's very kind of you! I will sent you an e-mail right now!

Thank you again!