On Thu, 3 Nov 2005 12:45:21 -0500, Daniel Mathalon
<[log in to unmask]> wrote:
>Dear Will Penny and others,
>
>In reviewing this e-mail response to Tobias
>Melcher, I had a couple of questions about PPI
>that perhaps someone can answer.
>
>1. In extracting the time series from the ROI
>using the VOI button, it asks whether we want
>adjusted or unadjusted data. I have read that we
>should choose to adjust for variables of NO
>INTEREST. In our case, we have used 6 motion
>parameters as covariates of no-interest in the
>model, and there are also separate sessions. It
>seems that VOI only lets us adjust for ONE effect
>in the model--i.e., there is no obvious way to
>choose the 6 motion parameters at once as the
>variables to be adjusted for. How does one get
>adjusted data, where the time series is adjusted
>for MULTIPLE variables of no interest?
I know this response isn't timely, but I've been looking into this myself
right now.
The way to adjust the data is to construct an F-contrast that subsumes all
the variables of interest and leaves out the ones you're not interested
in. Usually, as outlined in Will Penny's how-to textfile, you
pick "effects of interest". At least on my SPM2 setup, this will include
the motion regressors, which you don't want. So create an F-contrast
which excludes the motion regressors and the constant-per-session columns
(viz, the effect of session columns). Then adjust for this new F-contrast
instead of "effects of interest".
[snip]
I snipped the remainder of your questions since I'm not sure any answers
wouldn't be too little too late.
Cheers,
S
|