Since I'm assuming you're using a fixed effects analysis at the second level
to combine runs over subjects, in the case of a single run using the first
level cope is the same as the special case of a fixed effects model with 1
observation.  So you can use first level copes for the 1 run subjects and
any difference due to one subject having more runs than another will be
reflected in smaller variances for the subjects with more runs.  Make sure
to use FSL's FLAME1 for the group analysis to account for this


On Wed, Dec 2, 2009 at 12:57 PM, Agatha Lenartowicz <[log in to unmask]>wrote:

> Hello ~
> I have a group of subjects, most of whom have 6+ runs, but a few have just
> 1 run. Typically, I would conduct a second level analysis to combine across
> runs, and then a 3-rd level group analysis on these results. However,
> clearly I can not run a 2nd level analysis for subjects with one run. I
> appreciate that with an imbalanced set of data across subjects I may have
> issues at the group level, but I would explicitly like to see the effect of
> these subjects on my analysis.
> My question is - can I pass the cope.nii.gz results from the first level
> analysis (for those subjects with 1 run) directly into the third level?
> Intuitively this seems ok - since the PE values for 1 run are all there is
> for those subjects!
> Many thanks!
> Agatha