Dear Ofir,
The functionality you are referring to in the 'fMRI model specification'
module (Factorial design) is there for convenience only and indeed
assumes that the defined conditions all correspond to cells of a fully
balanced factorial design. This means you cannot use this functionality
if you have extra experimental conditions you want to model.
In this case, specify explicitly the contrasts of interest at the first
level and proceed to the second level with these contrast images (using
one-sample t-tests). If you need help to define the various contrasts,
you can use this:
spm_make_contrasts([2 2 2])
assuming the conditions are entered in the order defined in the help
text of the first-level Factorial design option.
Best regards,
Guillaume.
On 17/01/2019 20:06, Ofir Shany wrote:
> Hello SPM users,
>
> I have a 2X2X2 within-subject design. Ultimately my goal is to test main effects and interactions at the group-level.
> I understand that I can model the different factors at the 1st level design, and that SPM will automatically identify the conditions that comprise the different factors and compute the relevant F-contrasts for main effects and interactions.
> However, in addition to the 8 conditions that comprise the levels of the factorial design, I have an additional regressor which represents the appearance of an instruction phase throughout the task. As far as I know such an instruction phase should be modelled in the GLM even though it's not a part of the experimental conditions.
> Will the software recognize that this condition isn't a part of the factorial design, as it appears after the first 8 conditions I defined? Should this condition be defined where confound regressors are added – because it seems that it will not be convolved with the HRF if it's defined there; Or should such a factorial design only be defined in the 2nd level analysis?
>
> Thanks,
>
> Ofir
>
--
Guillaume Flandin, PhD
Wellcome Centre for Human Neuroimaging
UCL Queen Square Institute of Neurology
London WC1N 3BG
|