Hi,
This should be ok - you can run a separate second-level analysis for
each subject to combine across sessions; it's probably best to choose
the fixed-effects modelling option for this. See the FEAT manual for
more examples of higher-level FEAT models.
Cheers, Steve.
On 11 Aug 2007, at 12:55, Guido Biele wrote:
> Dear Steve, dear Stephane,
>
> thanks for your responses!
>
> I am aware that concatenating is generally discouraged. However, I
> fear that
> my data is more messy than my first email suggested: We were
> running an
> (event related) experiment where the conditions of interest are
> determined
> by participants decisions.
> As a result, some types of decisions (conditions) were made
> infrequently and
> are in addition scattered accross runs; lets say 0,4,10, and 11
> events in
> runs 1,2,3, and 4, repectively.
>
> Would you say that, for this kind of data, we get more reliable
> parameter
> estimetes by running a multi level analysis for each individual,
> compared to
> concatenating runs?
>
> best - guido
>
>
> On Fri, 10 Aug 2007 17:26:31 +0100, Steve Smith
> <[log in to unmask]> wrote:
>
>> That's correct - we strongly discourage temporal concatenation of
>> datasets across sessions for FEAT analysis.
>> Cheers.
>>
>>
>> On 10 Aug 2007, at 17:14, Stephane Jacobs wrote:
>>
>>> Hi Guido,
>>>
>>> If the only reason why you want to concatenate your runs is because
>>> you
>>> want to contrast conditions belonging to separate runs, I *think*
>>> (and
>>> please anybody correct me if I'm wrong!) that you don't need to do
>>> that.
>>> Instead, you could just model each run individually at the 1st
>>> level,
>>> and then contrast conditions as you wish at the 2nd level (for each
>>> subject) using cope images from the 1st level .feat directory as
>>> inputs.
>>> After that, you can average across subjects at the 3rd level.
>>>
>>> Hope this helps,
>>>
>>> Stephane
>>>
>>> Guido Biele wrote:
>>>> Hi,
>>>>
>>>> I am comparing two motion correction methods for concatenated
>>>> functional data (something I have
>>>> to do, because conditions to be contrasted are not always in the
>>>> same run).
>>>>
>>>> The methods are:
>>>> a) simply concatenating the functional data anf then using mcflirt.
>>>> b) first using mcflirt for seperate runs, then using mclirt to
>>>> register each run`s example_func to the
>>>> global example_func, and finally concatenating the two MAT_...
>>>> files to register each volume on the
>>>> global example_func.
>>>>
>>>> I have two questions in this context.
>>>> Does the second, rather complicated method make any sense to you,
>>>> or should the results be the
>>>> same as for the first, simpler method?
>>>> Can one combine the transformation paramteres (from the xz.par
>>>> files ) by simply adding them?
>>>>
>>>> Cheers, guido
>>>>
>>>>
>>>>
>>
>>
>> ---------------------------------------------------------------------
>> ---
>> ---
>> 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
------------------------------------------------------------------------
---
|