Dear Claudie,
with a 2x2x2 ANOVA within-subjects you have two options:
* use a flexible factorial design with four factors: subject, block,
speed and difficulty. Then specify 'main effect' of factor 1 and
'interaction' of factors [2 3 4]. This will require you to edit
config/spm_cfg_factorial_design.m at line 579 so that it reads:
fnums.num = [Inf 1];
* specify your question of interest at the first level and use several
second level models. Given that all of your factors have no more than
two levels, all of your questions can be defined with unidimensional (t)
tests at the first level and second level models will all be one sample
t-tests. You can use:
>> C = spm_make_contrasts([2 2 2])
to automatically obtain the contrast vectors to use at the first level.
As for the unusual error message you get, I don't know. It would be
useful to send me the SPM.mat corresponding to that analysis.
Best regards,
Guillaume.
On 31/12/15 08:44, GAILLARD Claudie wrote:
> Hi everybody,
>
>
>
> I am very grateful for all the helpful information on the spm email
> discussion list as well as on the website of the Wellcome Trust Centre
> for Neuroimaging.
>
>
>
> As I’m a full beginner in MRI analyses and in using Statistical
> Parametric Mapping, I am sorry if the asked questions are basic and will
> be very thankful for any advice you could give me.
>
>
>
> In fact, I have a 2 x 2 x 2 design regrouping 3 within-subject factors,
> with factor 1 consisting in two blocks (i.e. Block 1 vs. Block 2),
> factor 2 consisting in two levels of speed (i.e. slow vs. fast) and
> factor 3 consisting in two levels of difficulty (i.e. low vs. high).
>
>
>
> To test the contrasts of interest, I tried different way, the first one
> was to perform at the first-level analysis four one-sample t-tests on
> each subject (1) Block1_Slow in low difficulty>Fast in low difficulty,
> (2) Block2_Slow in low difficulty>Fast in low difficulty, (3)
> Block1_Slow in high difficulty>Fast in low difficulty, (4) Block2_Slow
> in high difficulty>Fast in low difficulty.
>
> Then in the second-level analyses, (1) four one-sample t-tests with all
> subjects (the same as previously for each subject independtly) and (2)
> two paired-sample t-tests to test the differential activation between
> block 1 and block 2 for each level of difficulty.
>
>
>
> However, I encountered a problem in the first stage of analyses (i.e.
> paired-sample t-tests). Indeed, Matlab displayed this error message :
>
>
>
> Index exceeds matrix dimensions.
>
>
>
> Error in spm_list (line 709)
>
> if ~isempty(TabDat.dat{i,5}), fw = 'Bold'; else fw = 'Normal'; end
>
>
>
> Error in spm_list (line 138)
>
> spm_list('Display',TabDat,hReg);
>
>
>
> Error while evaluating uicontrol Callback
>
>
>
>
>
>
>
> I have no idea how to manage this problem. Would you have any advice ?
>
>
>
> So far, I envisaged a full factorial design to perform 2 x 2 x 2
> within-subjects ANOVA in SPM12. However, I was stuck at the stage where
> it was needed to select the images for each cell. Indeed, I haven’t
> images regrouping all scans of each variable (i.e. cell) for all
> subjects. Would you have any recommendation how I could manage that ?
>
>
>
> I thank you a lot for any advice and recommendation you could give me
> and wish you still all the best for the New Year.
>
>
>
> Best regards,
>
>
>
> Claudie
>
>
>
--
Guillaume Flandin, PhD
Wellcome Trust Centre for Neuroimaging
University College London
12 Queen Square
London WC1N 3BG
|