Hello,
multiple regression means that you define equally regressors that may
explain your observation
(which is grey matter value/volume per voxel). You simply include your
neuropsychological variable,
and in addition the nuisance covariates (usually total grey matter, age and
gender) after selecting
that model type.
The contrast you are interested in is probably a positive or negative
correlation between
your observation and the first covariate: so in the CM define a t-contrast
with "1" or "-1" set
onto that regressor; the others should be set to zero or are padded
automatically.
Your neuropsychological measure should be suited for such a manoever (i. e.
needs to be a parametric scale).
HOWEVER, writing all this, I am not convinced that this study - if it is
not just a pilot-like test analysis
and much more subjects will follow - will give a meaningful result. 10
subjects are to few to be representative, given the large variance of cortex
anatomy.
IN ADDITION, you write you have used the "optimized" scripts etc. which to
my knowledge includes the creation of an
own template - 10 subjects are MUCH too few to create a template and is a
wrong strategy here. Instead of the
workarounds used in "optimized VBM" better switch to SPM5 and avoid any own
templates at this stage.
Alternatively, if you definitely only have 10 subjects, maybe better
restrict the analysis to a simple, old-style volumetry and measure
larger volumes as lobes which gives a better power than a voxelwise
analysis with few subjects.
best wishes,
Philipp
Max-Planck-Institute of Psychiatry
NMR Research Group
Kraepelinsr. 2-10
80804 Munich
Mail: [log in to unmask]
Phone: 0049-89-30622-413
|