Hi Buyean
> Hi Cyril,
> Is the regression analysis with cste (the cste modellig the mean
> activity) the same as the Simple regression (correlation) in the Basic
> models?
I don't remember well what's in SPM2 but I would say yes - what you
should get is a design matrix with one column for the cognitive scores
and one extra column created automatically that models the mean.
> If this is true, would you mind giving me a very general guideline for
> the following options?
> (I read the help file and other websites, but still I am not sure what
> I need to do with these options).
> Threshold masking: when should one use this option?
Threshold is use for instance in VBM. It allows to make sure you are not
imaging an other tissue type. In the case of dopamine receptors I
suppose there is no 'risk' of imaging smthg else so I would not use any
threshold but I may be wrong.
> explicitly mask images? Is this going to limit the voxel search to
> the brain regions in the mask?
> Is this the same as the small volume correction?
The idea behind is the same, this will limit the search space. Now in
terms of stats I think the computation is the the same. With explicit
mask you reduce the space and the correction for multiple comparison
will be based on this space (smaller than the whole brain but usually
much bigger than what you use with the SVC button). With the FWE (random
field) the stats are based on this overall space (shape - ie nb of
resels). With the SVC, the correction is based on the shape and size of
a particular (small) area you selected. (see Worsley et al. 1996 HBM 4).
I never tried but maybe (I say maybe because I'm not sure) if the mask
is a particular area, you will get the same (or similar) result as if
you were looking at the results in this area with the the SVC -
/Anyone to comment on this ??/
> Global calculation: when should one use this option?
In fMRI we don't use it anymore; sorry but I don't know for your
particular case. However as you have a cste in your model I guess this
is fine not to compute the globals.
Best
Cyril
> > Dear SPMers,
> > I am investigating if there is a significant relationship between
> the level
> > of dopamine receptors and one of the cognitive tasks.
> > There are twelve binding potential images (e.g., images of dopamine
> receptor
> > density) and 12 numbers for the cognitive task.
> > I will greatly appreciate if someone can guide me with this analysis.
> > (fowarding emails, any website, or any reading material that I can
> start with).
> > Based on my reading of numerous websites, I think I need to use the
> simple
> > regression, but there are many options in SPM2 and I am not sure which
> > option I need to choose.
> I'm not familiar with receptor binding but I'm quite confident with
> the approach you suggest.
> In SPM simply use the regression analysis with cste (the cste modellig
> the mean activity) and regress, ie find the the voxels in which you
> have a correlation, the cognitive test values. First enter the 12
> images then add 1 regressor and enter the scores. Estimate and look at
> the result :-) that's it.
>
|