You want to process the images so that the null hypothesis would be valid. Under the null hypothesis, there's no difference between the groups, so you assume they are the same. However, if you systematically treat one group differently from the other (by including a greater proportion of scans in the template), then you would/could be introducing differences.

In general, you should process the data in a way that is blind to the hypotheses that you intend to test.

Best regards,

-John

On 15 April 2015 at 11:26, Marieke <[log in to unmask]> wrote:

Hi,

I want to compare grey matter volume between a large group of patients (N= about 90 females) and smaller group of healthy controls (N=about 30 females). My question concerns the dartel template.

In a previous post (21 May 2012) I read that if you include all subjects, the template will be biased towards the larger patient group and this also biases normalization. A suggestion was to randomly select equal numbers from the patient and healthy control group.

I could select all the 30 healthy controls and select a subset of patients (30 that are matched for age and education). But, doesn't this also bias normalization because now the majority of the patients are not in the template?

Suggestions on how to solve this would be very welcome,

Best, Marieke