Hi - sounds like the reviewer is correct - because of (within-subject)
correlation between inputs to the second level in case 2, case 2 is
overestimating the true degrees of freedom and so, without covariance
correction, is not quite right. Case 1 is fine and is the easiest correct
thing to do.
Cheers, Steve.
On Mon, 2 Jan 2006, Helmut Laufs wrote:
> Dear list,
>
>
> Two scenarios:
>
> #1:
>
> First level: 11 subjects, two sessions per subject
> one contrast image created per subject including the effect of interest of both sessions, i.e. 11 con images
>
> 2nd level:
> SPM2, one sample t-test, "+1" contrast for these 11 images
>
> #2:
>
> First level: 11 subjects, two sessions per subject
> one contrast image created per SESSION including the effect of interest of each session separately, i.e. 22 con images
>
> 2nd level:
> SPM2, one sample t-test, "+1" contrast for these 22 images
>
> Question:
>
> I (a medic), assumed scenario #2 would have more power, i.e. in a manuscript I wrote
> "... resulting in higher significance (more degrees of freedom)".
>
> "My reviewer" now says:
>
> "If the authors think so they should think again! Was the covariance structure
> properly modeled?"
>
> What shall I reply? Where am I wrong? What about 'modelling the covariance structure properly'?
>
> Any help (as soon as possible :-) ) greatly appreciated!
>
> With best wishes for 2006,
>
> Helmut Laufs
--
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director, Oxford University FMRIB Centre
FMRIB, John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
|