Dear Jesper,
Thanks for the reply.
Actually, I have been doing my second level analysis using con***.img from the
first level , exactly as you suggested.
What happened was that I read that it is possible to use adj.meand fMRI summary
statistics in the RFX type of analysis and that the adj. meansfMRI is a way of
summarizing ones data.
I had no intentions of using the adj. means to do the stat. However I wanted to
explore what SPM99 does and how it does it.
Now, given the constraints that you mentioned in your reply, what's the purpose
of then having an option to collapse data in an fMRI exp. I mean, how does one
use the adj means data, and for what purpose , to what end?
Thanks again,
/Priyantha
Jesper Andersson wrote:
> Dear Priyantha,
>
> >dear all,
> >
> >I have been trying to use the adj means/fmri button to collapse my
> >fmri data with the intention of using the summary data in a second
> >level analysis.
> >However, as usual, i could not figure out how to do this!
> >i am asked for the epoch order within the session.. which is OK. let
> >us say i have 5 conditions within a sessions such that
> >RACBRDCER where R is rest A, B, D, E are task blocks and C is
> >another base line. For each block I have 6 functional volumes. I get
> >in to trouble when I start filling in the number of volumes per
> >condition.. it goes on and on and on... how come?
> >Can somebody please spend a little time and explain to me how to
> >do this properly?
> >
>
> I don't really know why you are having problems. Lets say you have selected
> 54 scans and enetered the string above (RACBRDCER), it should then be
> possible for you to specify 6 when asked for #scans/Epoch. If on the other
> hand you had selected e.g. 51 scans (e.g. because three first were dummy
> scans) you would then need to specify the length separately for each epoch,
> i.e. 3 6 6 6 6 6 6 6 6. Hence my only possible explanation for your problem
> is if the number of epochs you specify multiplied with the #scans/Epoch
> don't equal the no. of scans you selected.
>
> Having said that I would strongly recommend using con*.img from the first
> level analysis for your second level analysis rather than using the AdjMean
> function. You should create your contrasts at the first level testing them
> with a one-sample (or two sample if you look for group-by-contrast
> interactions) t-test at the second level. Taking the condition means to the
> second level enetering them into a PET type analysis potentially inflates
> your type I errors since you then implicitly assume that the inter-subject
> variablility in activation to one task tells you something about the
> inter-subject variablility in activation to another, which is pretty
> unlikely to be true.
>
> >Thanks a lot,
> >
>
> >
> >
> >Priyantha Herath MBBS
>
> Good luck Jesper
>
> Jesper Andersson
> Wellcome Dept. of Cognitive Neurology
> 12 Queen Square
> London WC1N 3BG
> phone: 44 171 833 7484
> fax: 44 171 813 1420
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|