Hi Cathy
Thank you for your answare.
> Contrast 2 allows the subjects with big condition effects on p to
> dominate the SPM whereas contrast 1 places even weights on each subject.
> How many of your subjects showed an effect of condition on p? The effect
> may just be coming from one or two subjects
I have two idea to estimate the effect of condition of subjects:
1. Plotting effect of interest under contrast 1 & 2 at same point.
2 Using subject specific contrasts such 1 -1 0 0 ...; 0 0 1 -1 0 0 ..;
...
Which is the correct (or better)?
I think that using contrast 2 and plotting effect of interest the
effect of condition on p will be shown while using contrast 1 the
effect of condition on paradigm will be shown. (?)
The condition effects of subject using contrast 1 and 2 at same point
based on plotting effect of interest are:
Subject: 1 2 3 4 5 6 7 8
contrast 1 - + * + + + + + (T=5.91)
contrast 2 - + * + + * + + (T=6.40)
(- : no effect, + : normal effect, * big effect)
Does it mean taht the subject 3 and 6 have big condition effect on p and
so they dominate the SPM?
Using subject specific contrasts the conclusion was that only
subject 3 has large (significant) respons at the same point.
(This result correlates with the plotted effect of interest under
contrast 1.)
By the way could you suggest any reference where thi factorial approach
type contrast were used?
Many thanks
Miklos
> Best wishes
> Cathy
>
>
> At 08:52 PM 4/29/2001 +0200, you wrote:
> >Dear SPMers,
> >
> >I am working on an analysis of a PET activation study (8 subj, 2 cond, 3
> >repl).
> >Multi-subject: conditions x subject interaction model was chosen
> >and two contrasts were used for generating SPM maps.
> >Contrast1: 1 -1 1 -1 1 -1 1 -1 1 -1 1 -1 1 -1 1 -1
> >Contrast2: p1 -p1 p2 -p2 p3 -p3 p4 -p4 5 -p5 p6 -p6 p7 -p7 p8 -p8
> >
> >Where pi the mean of an electro-physiological parameter of ith subject
> >(these parameters were measured during all PET scans)
> >These parameters were also used as indicators of the subjects in the
> >subject selection procedure.
> >
> >The result given by Contrast2 was better then Contrast1:
> >the t-values of cluster-maximums were higher and
> >the sizes of clusters were larger.
> >
> >My questions are:
> >
> >- Is the Conrast2 correct? (as factorial approach)
> >If it is:
> >- Can we say that the activation of these clusters have a parameter-dependent
> >augmentation?
> >
> >Many thanks in advance
> >
> > Miklos
> >
> >
>
|