Patrick
I think the answer is that the 'Plus 10' transformation you are using is
not 'linear' in the sense that you are expecting. If I understand corectly,
you might have started with a regressor
R: 0 0 0 5 5 5 5 0 0 0 0 0 8 8 8 8 8 0 0 0
to represent two box cars of heights 5 and 8 of lengths 4 and 5. Then you
constructed the regressor
R': 0 0 0 15 15 15 15 0 0 0 0 0 18 18 18 18 18 0 0 0
However R' is not a linear transformation of R. (E.g. the correlation of R
and R' is not 1 but 0.985.)
Ian
At 02:31 PM 23-01-06 +0100, Patrick Britz wrote:
>Dear SPM users,
>
>I am confused about the effect of linear transformation on my results:
>We have an event related design with three different (repeated) conditions.
>We also have a numeric value for every stimulus and want to determine the
>correlation between the bold response and the numeric value. We have
>modelled a user defined regressor for each subject in which we used a
>boxcar-function with the length of the event and with the numeric value of
>the stimulus as the height of the rectangle. This regressor was convolved
>with the hrf.
>
>My question is: If I transform the numeric value of every stimulus with the
>identical linear function (e.g. original value+10), the results change. This
>is confusing. As far as I know correlations are independent under linear
>transformations.
>
>Any help would be greatly appreciated!
>Patrick
>
>Dipl.-Psych. Patrick Britz
>Universität Trier
>Fachbereich I
>Universitätsring 15
>54286 Trier
>0651-201-2896
>[log in to unmask]
Ian Nimmo-Smith
MRC Cognition and Brain Sciences Unit
15 Chaucer Road
Cambridge UK
CB2 2EF
Tel +44 (0) 1223 355294 x 710
Fax +44 (0) 1223 359062
|