Hi,
There is no way of specifying group membership for any of the input files and melodic will treat all files the same.
The only exception from this rule is that the first input data set is being used to e.g. derive the mean image (used for display in the report only, though you can always specify a different bg image, and to call BET for masking) and estimating smoothness in the data. None of this should introduce consistent bias for any between-group analysis on the basis of the output...
There certainly is going to be an implicit bias in case where group sizes differ but that's independent of the order of subjects...
hth
Christian
On 30 Jan 2013, at 22:11, Chris Keown <[log in to unmask]> wrote:
> When working with different groups, say autism and a control, is it necessary to randomize the order of subjects in temporal concatenated ICA? I just want to avoid one group biasing the results.
>
> Thanks!
> chris
|