Thank you for the insight; that is very interesting. I'm performing tests of continuous independent variables over a single sample of normally developing children (ages uniformly distributed from 8 to 15 along with 50/50 gender split). The IV is, for example, age-normed cognitive scores falling in a normal distribution. Since there is only one group, I suppose creating my own template from all samples could introduce significant bias; especially if there are hidden sources of heterogeneity I'm unaware of. 

Thank you.

Best,
Chintan


On 11/09/2017 07:07 AM, Greve, Douglas wrote:

> It might capture specific anatomical features that are unique to your 
> data set. It can be tricky though as you need to make sure to include an 
> equal mix of all your groups. Eg, if you have 20 ADs and 10 controls, 
> you'd need to make an atlas with 10 ADs and 10 controls.