Christy,
That's interesting, but in practice not unexpected. I'd say that there are
two fundamental sources of inter-subject variability in the smoothness
estimates. One is the estimation error (Poline et al., 1995). The second is
time series model validity, which could vary systematically across subjects
due to inter-subject variability in hemodynamic responses. If there is
enough spatially smooth, (true) signal that is not modeled (such that it
ends up in the residual), this will tend to increase the estimated spatial
smoothness.
Whether or not you can ignore it would depend on the nature of the
application. For example, if you are doing multiple single subject studies,
such that you desire a p-value per subject, you could choose a conservative
smoothness for p-value generation. For example, for each study you could
choose the smoothness that would ensue from the applied kernel (in your case
FWHM = 12 mm).
Best,
Eric
1) Poline JB, Worsley KJ, Holmes AP, Frackowiak RSJ, Friston KJ (1995)
Estimating smoothness in statistical parametric maps: variability of p
values. Journal of Computer Assisted Tomography 19:788-796.
----- Original Message -----
From: "Christy Marshuetz" <[log in to unmask]>
To: <[log in to unmask]>
Sent: Thursday, August 18, 2005 1:06 PM
Subject: [SPM] same smoothing kernal, different est. smoothness
> Dear SPM,
>
> I have an elementary smoothing question that I couldn't find an answer
> to in the archives. I have 4 participants, all of whom started with
> voxel sizes that had been resampled to 4 x 4 x 4 mm during realignment.
> I applied a smoothing kernal of 12 mm FWHM and ran models on each
> participant's data. When I look at the results, at the bottom of the
> results page, each participant has a different estimated smoothness.
>
> Participant 1: estimated smoothness (mm) = 11.9, 12.3, 11.5 mm = 3.0,
> 3.1, 2.9 (voxels).
>
> Participant 2: estimated smoothness (mm) = 19.0, 20.0, 17.5 mm = 4.8,
> 5.0, 4.4 (voxels)
>
> Participant 3: estimated smoothness (mm) = 14.0, 14.4, 13.1 mm = 3.5,
> 3.6, 3.3 (voxels)
>
> Participant 4: estimated smoothness (mm) = 15.5, 16.1, 14.1 (mm) = 3.9,
> 4.0, 3.5 (voxels).
>
> 1) Is this normal? If so, can I just ignore it?
>
> 2) If not normal, assuming that most of the time I didn't make any
> typographical errors, does anyone have any idea what might account for
> these differences?
>
> Thanks,
> Christy
|