The finding can easily be explained by the fact that you are analysing
different pre-processed data. For the TBM, you are analysing Jacobian
maps, whereas for the VBM, you have Jacobian scaled grey matter maps.
Changes are that one of these forms of data will be better for your
data and question, but a priori, I do not know which. The choice is a
model comparison problem - in a similar way to F tests being model
comparisons (where the null hypothesis model is nested within the more
complex alternative hypothesis model). This involves comparing two
or more candidate models, and assessing which one is more accurate.
For example, see:
Goodman, S.N. "Toward evidence-based medical statistics. 2: The Bayes
factor". Annals of internal medicine 130(12):1005 (1999).
It is possible to decide which is better by assessing which approach
can more accurately separate your groups. This accuracy may be
determined in a number of ways. Cross-validation would be one way,
but measures such as Bayesian model evidence can also be used. For
this kind of problem, accuracy can not be compared within a generative
model setting, because different data are involved. However, it is
possible to compare models within a discriminative setting. See (for
example) the following paper:
K.J. Friston, C. Chu, Janaina Mourão-Miranda, Oliver Hulme, G. Rees,
W.D. Penny, and J. Ashburner. Bayesian decoding of brain images.
NeuroImage, 39(1):181-205, 2008.
Best regards,
-John
On 8 August 2011 08:39, shilpi modi <[log in to unmask]> wrote:
>
> Dear SPMer's,
>
> While performing morphometric analysis between two groups of subjects both using VBM and TBM, I am getting differences in the results in few of the regions, i.e. some regions that show gray matter atrophy in VBM are absent in the TBM analysis and vice-versa. How can I explain such finding?
>
> Thanks and regards,
>
> Shilpi Modi,
> Scientist C,
> NMR Research Centre,
> INMAS,
> Delhi,
> India
|