At the risk of repeating myself, I would recommend you look at the
example below, where you would ignore the between-subject factor so that
it then conforms exactly to your 2x3 within-subject design:
In the design you specified, you should probably use Independence: No
for factors age and song. Then you should select:
Main effect: 1
Interaction: [2 3]
(instead of Main effect: 1, Main effect: 2, Main effect: 3).
PS: the naming of the contrast images you bring to the second level seem
inconsistent across subject: this is entirely possible depending on how
you analysed the data at the first level but I'm just mentioning this in
case you accidentally didn't select the right files.
On 24/06/2019 20:22, Payal Arya wrote:
> Dear Guillaume
> Thank you for your reply.
> I set up a second level design. I have attached my design matrix. I am
> not sure if it is correct or not. It is a second level design, I had set
> up in the following way". 16 subjects in total.
> Factor2:age(Three levels)
> Factor3:Song(three levels)
> age has three level-55,66 and 90 dph.Song has three level:TUT1,TUT2,CON
> I used 1st level contrast images For example for subject1: at 55 dph I
> had three con images-TUT1>Rest,TUT2>Rest and CON>Rest.
> and 6 more con images from other two age for the same subject.
> I tried to use contrasts according to SPM_make_contrasts([3 3]), I
> always get invalid contrasts error.
> I am not sure if i am defining the contrasts correctly in this case.
> I do not see subject columns in the design matrix, Is this correct?
> Please could you help me in identifying the problem.
> Thank you
> On Mon, Jun 24, 2019 at 7:10 AM Flandin, Guillaume <[log in to unmask]
> <mailto:[log in to unmask]>> wrote:
> Dear Payal,
> With within-subject factors only, you could indeed use the "pooled
> error" approach, implemented with a flexible factorial design with three
> factors: subject, age and song, and test for main effects of age, main
> effect of song and age x song interaction, with F-contrasts given by:
> spm_make_contrasts([3 3])
> Alternatively you can use a "partitioned error" approach as described
> here and in Martyn McFarquhar's paper cited within:
> Best regards,
> On 21/06/2019 18:16, Payal Arya wrote:
> > Hi Spm users
> > after going through the archives related to flexible factorial design
> > for within subject factors. I have few questions , i was hoping i
> > get answers for them.
> > I have two factors which are within subject- *age* of the birds (cute
> > little songbirds) at which they are scanned and *song* type played
> > during scanning. Songs have three levels and age has three levels.
> > So I only have one group of 16 birds which are scanned at
> different age
> > while listening to different songs.
> > I made single subject design at 1st level at every age when they were
> > scanned.
> > question 1:
> > Now can i set up a flexible factorial design with age as one within
> > factor and song as second within factor and use contrasts from ist
> > level- song1>song2 or song2>song1?
> > question 2:
> > will the results of main effect of stimulus and main effect of age be
> > valid ?
> > questions3:
> > also is the interaction between two factors valid in such a case?
> > Thank you for your help
> > Payal
> Guillaume Flandin, PhD
> Wellcome Centre for Human Neuroimaging
> UCL Queen Square Institute of Neurology
> London WC1N 3BG
Guillaume Flandin, PhD
Wellcome Centre for Human Neuroimaging
UCL Queen Square Institute of Neurology
London WC1N 3BG