Dear Jonathan
I thank you very much for your help. The reference you give me was very
helpful.
To make sure that I understand the solution of the problem correctly:
- in general the regressors should not differ across sessions. If for
example the applied force, do not change across measurements, then I can
introduce this value as a regressor in the design matrix. If the force
differ across sessions than I should not use it.
- In the case of a training I have in addition the problem that if the
training has an effect on the regressor, for example on the force, by
introducing it in the design matrix I will maybe remove some of the real
training effect.
Question: I am sure that the applied force change after the training. But
since before each session I measure the maximal force, would it be correct
to normalize the force, by dividing it through the maximal Force, and to use
the normalize force as regressor? This of course only if the normalize force
do not differ between sessions!
Thanks a lot!
Natalia
-----Ursprüngliche Nachricht-----
Von: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]] Im
Auftrag von Jonathan Peelle
Gesendet: Freitag, 16. Dezember 2011 00:07
An: [log in to unmask]
Betreff: Re: [SPM] Regressors of no interest
Dear Natalia,
> I am analyzing an fMRI-training study we performed on stroke patients
> using a motor task. During the task I recorded several parameters like
> for example the force applied by the patients to perform the task and
> the range of movement. The patients of the study were measured twice
> before the training and after the training. Since the task performance
> of the patients is better after the training, I would like to
> introduce the aforementioned parameters (mean force per scan and mean
> range of motion per scan), as regressors of no interest at single
> subject level. The aim is to ensure that the observed brain activation
> after the training is due to the training itself and not because the
> participants performed the task better. Is it correct to use such
> parameters (force, range of motion) as regressors of no interest or not?
In general, including some of these additional parameters as regressors (of
no interest) seems like a sensible thing to do.
However, it gets tricky if they differ on one of these parameters across
sessions that you are interested in comparing. If they actually differ on a
regressor, then you probably don't want to include it. This is explained in
more detail in the very helpful Miller & Chapman paper:
Miller GA, Chapman JP (2001) Misunderstanding analysis of covariance.
Journal of Abnormal Psychology 110:40-48.
In which case, you will be left with pre/post training being confounded with
some of your other measures, and just have to live with that being a
possible confound. However, you may be able to get some information by
doing correlations across subjects—e.g., do participants who perform the
task better activate a brain region more or less than participants who
perform it worse. If not, that might also suggest that the differences you
see are not due to simple task performance.
Hope this helps!
Best regards,
Jonathan
--
Dr. Jonathan Peelle
Department of Neurology
University of Pennsylvania
3 West Gates
3400 Spruce Street
Philadelphia, PA 19104
USA
http://jonathanpeelle.net/
|