Hi Maren,

In addition to the case of variances not assumed to be the same as Jeanette commented, this setting can also be used in randomise to define exchangeability groups/blocks, which can be used with certain multivariate or longitudinal designs. So, it is an useful setting. There will be some documentation about this soon I hope.

All the best,
Anderson

2013/5/22 Jeanette Mumford <[log in to unmask]>

Hi,

In my experience it may only be an advantage for very large sample sizes. Otherwise, since it is defining groups in which separate between-subject variances will be estimated, these estimates have lower df and may be highly variable. Using one group only gives a single between-subject variance estimate, but it should be a better estimate.

Cheers,
Jeanette

On Wed, May 22, 2013 at 9:12 AM, Maren Strenziok <[log in to unmask]> wrote:

Hi,
I was wondering in which case it is appropriate to use the Group column in the Glm that is left from the first EV? I ask because I thought it might be usful to set up a design with independent groups, which does not seem to be the case according to the FSL Wiki.

Maren