On Wed, Sep 25, 2013 at 5:52 AM, Paret, Christian <[log in to unmask]> wrote:

To add on the question by Jaclyn:

For which type of statistical comparison does the orthogonality between a regressor and its parametric modulator qualifiy:

1.       Is it allowed to contrast parametric regressors of different conditions (say, in a with-in subject design in which subjects are presented aversive pictures, a contrast between the paramod of a “negative-viewing” condition and the paramod of a “neutral-viewing” condition)?


Yes. This means that you are comparing the slopes of the PM terms of two conditions. This is similar to paired t-test.
 

2.       Is it valid to contrast the parametric modulator of condition 1 with the (main) regressor of the same condition?

No. The slope of the PM has no expected relation to the main effect. If you were to compare them, it would be equivalent to asking if B is different A in Y=Ax+B, where A is the slope and B is the Y intercept.
 

Thanks in advance!

Chris

 

Von: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]] Im Auftrag von Chris Madan
Gesendet: Tuesday, September 24, 2013 8:45 PM
An: [log in to unmask]
Betreff: Re: [SPM] Orthogonalization of regressors in SPM8

 

I also can only think of parametric modulators.


~ Chris

 

On Tue, Sep 24, 2013 at 12:33 PM, Torben Lund <[log in to unmask]> wrote:

I can only think of parametric modulation, and the build in high pass filter, but otherwise not.

 

Best

Torben



Sendt fra min iPhone


Den 24/09/2013 kl. 19.40 skrev Jaclyn Hennessey Ford <[log in to unmask]>:

Good afternoon!

We are in the process of responding to a reviewer who asked about orthogonalization of regressors in SPM8. He/She believed that the default of SPM8 is to orthogonalize regressors with respect to the order that they were entered into the model, making the order very important for interpretation.

From our research online, it appears that this is not the case except in the case of parametric modulators, as it explained on this wiki: http://imaging.mrc-cbu.cam.ac.uk/imaging/ParametricModulations. However, before responding to the reviewers, we wanted to ask you, the experts, if you knew of any other situations where regressors might be orthogonalized by default, or if there are any other factors that needed to be taken into account in our response.

Thank you so much,
Jaclyn