On Thu, 17 Feb 2005 13:14:55 -0600, Glabus, Michael <[log in to unmask]>
wrote:

Hi, Mike.

>I would appreciate some guidance from list experts on this issue: implicit
>scaling (session specific).
>
>
>
>We conducted an activation (A)-rest (R) fingertap experiment (right hand)
>with 8 sessions: ARARARAR. Each session collected 10 EPI BOLD scans, TR= 3
>seconds.

Experiments with different conditions in different sessions are tricky.

>We conducted a standard (SPM2) fMRI analysis: entering the data as 8
>sessions; canonical HRF basis function; HPF=128 seconds; Serial
correlation

If each run is 10*3 = 30 s, why use a HPF with a cutoff of 128 s?

>(AR1); global calculation=mean voxel value; Grand mean scaling=session
>specific; and we tried this with and without global normalization with no
>major difference in results.
>
>
>
>The results are attached (3 figures). The top figure shows the result of
the
>tap v rest conditions; the second, the tap v rest session means; the third
>the combination of both.

I can't tell exactly what the regressors are.

Are the regressors
(1) the effect of convolving entire blocks (which are entire sessions)
with the HRF (for each type, A and R), and
(2) session effects?

That's what it looks like to me (except that I would have thought the
rightmost session effect regressors to be only 1 or 0, whereas in the
grayscale schematic they don't look perfectly white in the light blocks).

If so, the problem as I see it is that if you take a boxcar representing
one of the blocks---which is also one of the sessions---and convolve it
with the (canonical) HRF, you'll get "almost" another boxcar back.  Not
quite, because there is some lead-in and lead-out at the edges of the
block.  (If you take a really long block, longer than the characteristic
time of the HRF, and convolve it with the HRF, this would be much clearer.)

That means for any given session, the regressor formed from the session
(convolved with the HRF) isn't all that dissimilar from the regressor
equal to the block effect.

That's a problem, because then the regressors are highly confounded.  If
you conduct t-tests, you "miss out" on variance.  A good reference in the
context of neuroimaging is Andrade et al., "Ambiguous Results in
Functional Neuroimaging Data Analysis Due to Covariate Correlation,"
_NeuroImage_ 10, 483-6 (1999).  (This is also mentioned in stats books,
e.g. Netter et al.'s book on linear models, though you have to hunt to
find it.)

And to the extent the regressors are not the same, you're just measuring
the "edge effect" of the lead-in and -out of convolving with the
hemodynamic response, which isn't right from a modeling point of view
(IMO).

>What appears to be happening is that the signal is being removed by the
>sessions specific scaling, as the first contrast shows no activity in left
>sensorimotor area and right cerebellum, while this activity *is* present
in
>the modeled session specific global confound! Note the large T-values
>associated with this analysis and the one where the main conditions and
>session specific confounds are included.
>
>
>
>Can anyone explain why this is happening, and should we be worried about
>this in analyses that do not produce such robust activations?
>
>
>
>Regards - MFG
>
>