Dear Tom,
At 15:26 08/10/98 +0100, you wrote:
| I have only recently been introduced to SPM and have
| been successful in conducting statistical experiments which analyze the
| contrasts between groups or subjects. However, I wish to conduct an
| experiment which evaluates the statistically significant similarities
| between 36 different subjects, all scanned once.
Nice thought! ...but unfortunately SPM can't do that!
--------------------
SPM is predicated on a Statistical hypothesis testing approach, which seeks
evidence against a point hypothesis of no change. If the data (surmised by
a sufficient statistic) is highly unlikely to have arisen under this null
hypothesis, then we assert that there's significant evidence against it.
Since neuroimaging data is continuous, the null hypothesis of no change is
never true - there will be some small change. Given enough data one can
always reject the null hypothesis. The approach developed for situations
when one has scant data, and must protect against false assertions based on
this data. It's then implicit that anything significant corresponds to a
change that's important. (This may not be true for single-subject fMRI with
many replications!)
Under this framework you cannot prove the null hypothesis. In general, all
you can do to support a non-significant result is assess the power or
Type-II error (error of not detecting a change) for the smallest change
you're interested in. Because of the difficulty in doing this for images,
little progress has been made in applying these power techniques to
NeuroImaging. One approach that might be worth looking at is:
Van Horn et al. (1998)
"Mapping Voxel-Based Statistical Power on Parametric Images"
NeuroImage, 7, 97-107.
--------------------
Turning to your particular situation, you have 36 images you want to
examine for similarities. In this situation I would recommend a
multivariate investigative technique, such as PCA.
Hope this helps,
-andrew
+- Dr Andrew Holmes [log in to unmask]
| ___ __ ___ Robertson Centre for Biostatistics |
| ( ,) / _)( ,) Boyd Orr Building, |
| ) \( (_ ) ,\ University Avenue, |
| (_)\_)\__)(___/ Glasgow. G12 8QQ Scotland, UK. |
+----------------------------------------- http://www.rcb.gla.ac.uk/ -+
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|