Hi Yannis,
thanks for reply. I have asked around an am pretty certain that the
group means should not be modelled in comparisons of contrast images,
as these are difference terms that can be directly compared as
differences in group means have already been modelled (I think!?,
although I can see how group means within difference terms could also
be modelled). Also, SPM5/8 gives the default option of not modelling
constant columns for this reason. I may be able to use SPM8 to read my
SPM2 files and find a way that way. Thanks anyways.
Garret
> Hi Garret,
> Not sure if I understand correctly your question, and I am not
> familiar with SPM2 (I use SPM5/8). If you obtain your contrast images
> at 1st level analysis (i.e withon subject, contrasting one condition
> to another), then at second level analysis, when you ask questions
> about a group of participants (or compare groups), I think you should
> expect to get the constant columns. Now they model the group mean(s).
>
> Yannis
>
> Sent from my iPhone
>
> On 22 Apr 2010, at 09:26, "Garret O'Connell"
> <[log in to unmask]> wrote:
>
>> Hi friendly SPM users,
>>
>> I want to run an ANCOVA in SPM2 to control for gender across 2 groups.
>>
>> When I try, my design matrix has a constant mean column, which my
>> supervisors tell me should not be there because "if the input (as we
>> have) is a contrast image i.e. a subtraction of conditions, then
>> they already have effectively removed differences in the mean
>> activity per subject at the first level (they are 'difference
>> scores'), so we don't need to add the constant term at the second
>> level".
>>
>> So my design matrix has 4 columns when it should only have, 3 due to
>> the modelling of the group means.
>>
>> Can anyone provide input? Thanks for your time.
>>
>> Garret (Edinburgh)
>
>
--
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.
|