Dear Estephan,
This has been reported by one other person too, although I
have not been able to replicate it and track it down. In the
worst case the outlier script is incorrectly identifying these
timepoints and hence removing them without needing to.
However, even in this worst case there is *no bias* involved.
You simply lose two timepoints, which typically has almost
no effect in an average analysis, since out of 122 volumes,
the loss of 2 is normally inconsequential. So it is safe to
continue with this, but I will have another look at the script
and see if I can find a problem.
All the best,
Mark
On 14 Apr 2010, at 04:36, Estephan Moana wrote:
> Hello all, I've been using fsl_motion_outliers to remove motion
> peaks, and noticed that in most of my functional runs the reference
> functional volume is tagged as an outlier. My runs have 122 TRs, and
> TRs #61 and 62 are almost always being tagged as outliers, even for
> those runs that had minimal motion detected by McFLIRT (mean
> displacement peak of 0.3mm for absolute motion, for example).
>
> Is it ok for me to include the confound EVs matrix generated by
> fsl_motion_outliers for all functional runs despite minimal motion
> detected, or am I at risk of introducing bias as TRs #61 and 62 are
> almost always being "removed" from further analysis?
>
> Thanks for any insight.
>
> Estephan
>
|