Hi - yes, you're right - after many debates we've mostly been
convinced (thanks Jesper ;-) that the right second-level within-
subject cross-session modelling is FE - hence this recommendation is
changing for the new release...sorry should have made that
clearer...the new FEAT manual will explain the logic.....
Cheers, Steve.
from the new manual:
"If you are carrying out a mid-level analysis (e.g., cross-sessions)
and will be feeding this into an even higher-level analysis (e.g.,
cross-subjects), then you should not use the FLAME 1+2 option, as it
is not possible for FLAME to know in advance of the highest-level
analysis what voxels will ultimately be near threshold. In fact, in
such situations, you should probably use FE (fixed effects) for the
mid-level analysis, with a separate FE analysis for each subject.
This in effect treats the multiple first-level sessions (for each
subject) as if they were one long session, and ignores the session-
session variability (which arguably is not of interest, and which you
never model explicitly if you only took just one session for each
subject). This approach also avoids the problem of having to pool
second-level variance across subjects, in the case where you don't
have many sessions for each subject. Finally, it does not seem to
make much sense that a ME cross-session analysis leaves you less
certain about the combined within-subject-across-session results than
if you have just analysed one of the first-level sessions - hence FE
makes more sense."
On 8 Aug 2007, at 16:20, James Porter wrote:
> Hello-
>
> I'm a little confused by this answer. I was under the impression
> that using
> FE analyses to combine single-subject sessions was a dead-end as
> far as
> group analyses go. The FEAT Details web page states,
>
> "Also note that you should probably not feed up second-level
> FE analyses to higher-level analyses; you should use a full
> mixed-effects analysis at all levels."
>
> Is this not counter to what is written below, or am I misreading
> something?
>
> Thanks,
>
> Jim Porter
> TRiCAM Lab Coordinator
> Elliott Hall N437
> 612.624.3892
> www.psych.umn.edu/research/tricam
>
>
> On Wed, 8 Aug 2007 07:41:19 +0100, Steve Smith
> <[log in to unmask]> wrote:
>
>> Hi - I would combine across session within-subject using FE for each
>> subject separately at second level and then feed this up into a
>> cross-
>> subject analysis.
>> Cheers, Steve.
>>
>>
>> On 8 Aug 2007, at 02:39, Matthew Hoptman wrote:
>>
>>> Hi all,
>>> I'd like to set up an analysis in which I do a cross group
>>> comparison on
>>> conditions from two different tasks collected in the same scanning
>>> session
>>> (but in different runs). I see how to do paired t-tests, but this
>>> is sort
>>> of like doing a repeated measures ANOVA with a grouping variable,
>>> and I
>>> can't seem to get it to work out. When I set up the paired aspect,
>>> the
>>> groups selector seems to default to a single column. That makes it
>>> hard to
>>> do the grouping part.
>>>
>>> I guess what I'm thinking is:
>>>
>>> Group Patients Controls Condition S1 S2 S3
>>> 1 1 0 1 1 0 0
>>> 1 1 0 1 0 1 0
>>> 1 1 0 1 0 0 1
>>> 1 1 0 2 1 0 0
>>> 1 1 0 2 0 1 0
>>> 1 1 0 2 0 0 1
>>> 2 0 1 1 1 0 0
>>> etc.
>>>
>>> Any ideas?
>>> Thanks,
>>> Matt
>>
>>
>> ---------------------------------------------------------------------
>> ---
>> ---
>> Stephen M. Smith, Professor of Biomedical Engineering
>> Associate Director, Oxford University FMRIB Centre
>>
>> FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
>> +44 (0) 1865 222726 (fax 222717)
>> [log in to unmask] http://www.fmrib.ox.ac.uk/~steve
>> ---------------------------------------------------------------------
>> ---
>> ---
>> =====================================================================
>> ====
------------------------------------------------------------------------
---
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director, Oxford University FMRIB Centre
FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
------------------------------------------------------------------------
---
|