Thanks Eugene! So how would I use fslmaths at this final stage to
determine that RvsG contrast is a positive activation that I am
contrasting between groups? I hope this make sense and thanks so much!
Eugene Duff wrote:
> Hi Vina -
>
> 2009/5/20 Vina Goghari <[log in to unmask] <mailto:[log in to unmask]>>
>
> Hi,
> I guess this is where I get confused with what to do in my
> situation. I have a number of subjects in three groups each - each
> subjects three sessions of the same task.
>
> I have for example 2 regressors (red condition and green
> condition), so I set up the analysis for every session and then I
> combine across sessions for every subject:
> R 1 0
> G 0 1
> RvsG 1 -1
>
> Then I want to look at differences between the 3 groups for the
> subjects for cope 3 (RvsG):
> Group 1 > Group 2&3: 1 1 -2
> Group 1 > Group 2: 1 -1 0
>
> Therefore I can't mask the higher level analyses to look at
> positive activations at the more basic regressor analyses.
>
> Am I doing it correctly? And in this case would I use fslmaths to
> see if the contrasts of the differences > 0 for the higher level
> analysis?
>
>
> Yes, that seems OK (did you make an error writing down the G1 > G2&3
> contrast?). I would probably recommend also looking separately at
> differences in R across groups, and the same for G, because with only
> your contrasts you won't have any indication whether a change in one
> or the other or both is causing any effects in the contrasts of their
> differences. Also note that unless groups 2 & 3 have an association
> that makes them a natural group, results of the G1 > G2&3 contrasts
> will be a bit ambiguous. You won't know whether just one or both of
> G2 & G3 is different from G1.
> Best,
> Eugene
>
>
>
>
> Thanks!
>
>
> Eugene Duff wrote:
>
> Hi Vina
>
> 2009/5/18 Vina Goghari <[log in to unmask]
> <mailto:[log in to unmask]> <mailto:[log in to unmask]
> <mailto:[log in to unmask]>>>
>
> Yes that makes complete sense. However, it does not work
> for how
> I've set up my contrasts and the effects I'm looking for - are
> there any other ways?
>
> I'm not sure how your situation is different? You may be able
> to use fslmaths to mask your final thresh_zstat images.
> Eugene
>
>
>
> Thanks!!
>
> Eugene Duff wrote:
>
> If you want to limit you results to regions where, say, the
> first regressor was associated with a positive
> response, you
> can use contrast masking. Select contrast masking in the
> post-stats section and input the contrast where you
> tested for
> a positive response in that condition (e.g. the 1 0
> contrast).
> Selecting "Mask using z>0.. " makes the masking less
> severe,
> showing regions where the fit of the regressor was
> positive,
> but not necessarily significant.
> I hope that makes some sense!
> Eugene
>
> 2009/5/18 Vina Goghari <[log in to unmask]
> <mailto:[log in to unmask]>
> <mailto:[log in to unmask] <mailto:[log in to unmask]>>
> <mailto:[log in to unmask] <mailto:[log in to unmask]>
>
> <mailto:[log in to unmask] <mailto:[log in to unmask]>>>>
>
>
> Thanks for your response. What is the way to ask for
> activations > 0?
>
> Thanks!
>
> tEugene Duff wrote:
>
> Hi Vina,
>
> 2009/5/18 Vina Goghari <[log in to unmask]
> <mailto:[log in to unmask]>
> <mailto:[log in to unmask] <mailto:[log in to unmask]>>
> <mailto:[log in to unmask]
> <mailto:[log in to unmask]> <mailto:[log in to unmask]
> <mailto:[log in to unmask]>>>
> <mailto:[log in to unmask] <mailto:[log in to unmask]>
> <mailto:[log in to unmask] <mailto:[log in to unmask]>>
>
> <mailto:[log in to unmask]
> <mailto:[log in to unmask]> <mailto:[log in to unmask]
> <mailto:[log in to unmask]>>>>>
>
>
>
> Hi,
> A couple of quick questions about when setting up
> contrasts:
>
> (1) If you have two regressors that at a
> lower level you
> set up:
> R1 0 1
> R2 1 0
>
> And then you contrast: -1 1. Do you still
> have to be
> worried that
> the contrast could stem from a difference
> that is below
> zero for
> both regressors?
>
> If I understand you correctly, then yes, this
> contrasts
> would
> identify differences in responses even if they
> are all
> negative responses (deactivations). (2)
> When you have three groups can you set up your higher level
> contrast this way to sum 0 (1 1 -2)?
>
> This contrast will find regions where the responses
> modelled
> by the first two evs are significantly greater
> than those
> modelled by the third EV. These regions may not
> necessarily
> remain signficant when only one of the first two
> evs is
> contrasted with the third.
> Best
> Eugene
> Thanks!
>
>
>
>
> --
> Eugene Duff
>
> FMRIB Centre,
> University of Oxford
> John Radcliffe Hospital, Headington OX3 9DU
> Oxford UK
>
> Ph: +44 (0) 1865 222 739 Fax: +44 (0) 1865 222 717
>
> --
>
>
>
>
> --
> Eugene Duff
>
> FMRIB Centre,
> University of Oxford
> John Radcliffe Hospital, Headington OX3 9DU Oxford UK
>
> Ph: +44 (0) 1865 222 739 Fax: +44 (0) 1865 222 717
>
> --
>
>
>
>
> --
>
> Eugene Duff
>
> FMRIB Centre,
> University of Oxford
> John Radcliffe Hospital, Headington OX3 9DU Oxford UK
>
> Ph: +44 (0) 1865 222 739 Fax: +44 (0) 1865 222 717
>
> --
>
>
>
>
> --
>
> Eugene Duff
>
> FMRIB Centre,
> University of Oxford
> John Radcliffe Hospital, Headington OX3 9DU Oxford UK
>
> Ph: +44 (0) 1865 222 739 Fax: +44 (0) 1865 222 717
>
> --
|