Hi Zhiqiang,
I'm afraid I'm not sure I fully understand you, but I think your
question is about covariate centering. I don't think user-specified
covariates are convolved with the HRF, so I don't think this is the
issue here. Quite often, centering doesn't affect the statistical
tests you are interested in anyway (even though the design matrix
picture might look bad). For example if your design matrix contains a
constant column (as I believe fMRI ones will, but I'm not sure) then
centering a covariate will just change the beta for the constant term,
which isn't usually included in contrasts of interest.
In some cases, e.g. (second-level) ANOVA designs with covariate-factor
interactions, centering can be important, and perhaps a little
complicated, but I don't think you are talking about such a case.
It is probably safe for you to simply subtract the mean of your
covariate, e.g.
signal = signal - mean(signal);
This should make the design matrix look nicer, but probably won't
change the results you are interested in.
I hope that helps,
Ged.
> I wanted to use a extrinsic matrix(signal) from somewhere else. the
> value of this signal was ranged from 900 to 1100. I employed 2
> ways to do.
>
> 1. put the matrix into "regressor" column, let the trail/conditions
> as "0". the producted matrix looked rather good. 2. made a casual
> SPM.mat, opened SPM.xX.X, replace the first column with my extrix
> matrix. but this producted matrix looked bad representing with a
> whole white column.
>
> I looked into the natural matrix of SPM.xX.X in SPM.mat, it ranged
> with positive and negative number, I know it came from the original
> value convolved with hrf, but I don't know how to simulate my
> signal as the model of natural SPM.xX.X. what the range? How to
> define the "0"? what's the proportion above and under "0"?
>
> Any answer is welcome! Thank you very much!
>
> Zhiqiang Zhang Lab of Neuroimage Nanjing Jinling Hos. Medical
> School. Nanjing University. China
>
> ------------------------------------------------------------------------
> ÑÅ»¢Ãâ·ÑÓÊÏä-3.5GÈÝÁ¿£¬20M¸½¼þ <http://cn.mail.yahoo.com/>
|