Ciao Cristina,

>
> I'm dealing with subjects that tend to produce large, frequent
> movents in all planes. They also have very tiny heads.
Nice setup, try giving them some wine before they go in the scanner.

> So far, I have decided which datapoints were unusable based on
> visual inspection of each run and the output of 3dVolreg, and
> censored them while running 3dDeconvolve.
That is OK. I am assuming you're also including the parameters from 3dVolreg as regressors of no interest.

>
> Getting rid of those datapoints (I mean, physically eliminating
> them from my bricks) would be a total nightmare and also
> methodologically incorrect, as I understand.
Very very bad idea. Do not do this.
>
> I was wondering though if it would be a good idea to replace
> the nightmarish subbricks with something like a non-offensive
> datapoint, or a mean image of some sort. The plugged-in images
> would still be ignored in Deconvolve. Do you see any advantages
> in following this procedure? How would I go about it?
Why do you want to do that? You don't need to put fake data in there. If you are censoring these points then you don't need to worry about their value in the analysis.

If these censored points look like spikes AND they are caused by motion AND you just want to make the time series look pretty for display purposes, then you can use 3dDespike which would do something along the lines of what you were thinking of. In your analysis this action is a matter of style, which might count for a lot when you work in Milano.

>
> thank you so much for your great work
>
> cristina
>
cheers,
z