Dear folks,

We have encounter this wired distribution of subjects as below. The attached picture is a frequency of count of certain behavioral in a given time(did ln(x+1) transformation). We used the count of behavior to correlate with their gray matter volume and their functional MRI task data. We got results as below:

1. without subjects who have zero count of the behavior, we got negative correlation between the count of behavioral and gray matter volume as well as between the count of behavior and functional data in a same regions. This result is very good and makes a lot of sense.

2. However, when we added in those subjects with zero count of behavior, those results disappeared.

3. we also tried to separate subjects into two groups: zero count group and non-zero count group. But, there was no group difference in either VBM or functional data.

Could you please instruct us on how to deal with problem like this? Your comments or suggestion are greatly appreciated!

--

Best Regards

Xue, Feng Ph.D.

Psychology Department
University of Southern California
Los Angeles CA, 90089
Mobile: +1 213-249-1040
=========================================
XFSL, script that facilitate your neuroimaging

data analysis: http://xfsl.fmri.cn