>Yanjun: I presume you have a set of single PET studies across a single
group of subjects, each with one or more ratings of some clinical measures.
>Then: in SPM terms, a "single-subject" model is appropriate, since-- in
the world of GLM, it seems-- single images across individuals correlated
with a variable is equivalent to multiple measures of a single subject
correlated to teh same variable [assuming it would *vary* over time].
>In general, you are absolutely right, a contrast of parameter estimates
must sum to zero-- but this is true for the usual case, that each parameter
estimate cannot be evaluated uniquely on its own-- such as the parameter
estimate for condition A or B alone in an activation study. They can only
be estimated relative to each other-- and then the contrasting weights must
cancel out.
>However, in the case of a covariate of interest with a value for every
scan, such as yours, the parameter of interest is fully determined by your
data. so-- if you had only one covariate of interest, your "contrasts"
would be {1} and {-1} for +ve and -ve correlations.
>If you have 3 covariates, then you enter zeros for all but the one you
want to know about: {0 0 1} is +ve correlations for the third variable,
etc. In this case, you can enter actual "contrasts" of variables, which
then represent tests of interactions of those terms. The interpretation of
such "interaction" SPMs is over my head, at this point, however. I suggest
you query this line again when and if you get to that point.
>
>Note that the values in the SPM{t} for any of these correlation tests will
not be correlation coefficients; extracting those takes another step or
two-- and I would be very grateful to those out there who can specify those
steps to us at this point.
>
>At 12:27 PM 11/01/2000 , you wrote:
>>Dear SPMers,
>>
>>I'm doing some correlation analysis between PET images and clinic data. I
>>select " Single subject: Covariates only". According to the content in Page
>>22 in Ch3.pdf of notes97, some contrasts with only +1 or -1 associated to
>>interested covariates should be proper selection. While SPM can only accept
>>contrasts summing up to zero. So, how to create the contrast for such case?
>>Thanks a lot.
>>
>>Sincerely,
>>Yanjun
>>
********************************************************
Christopher Gottschalk, MD *
Assistant Professor of Neurology & Psychiatry *
Yale School of Medicine *
*
Mailing Adress: *
VAMC [116-A] tel [203] 932-5711 x4329*
950 Campbell Avenue FAX 937-4791*
West Haven, CT 06516 *
********************************************************
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|