Hi Clark,
This reviewer is rather confused...
In fact the latest release of FSL simply uses the following reference
regarding inference using Gaussian field theory, as it brings all of
Keith's relevant work together fairly nicely:
K.J. Worsley. Statistical analysis of activation images. Ch 14, in
Functional MRI: An Introduction to Methods, eds. P. Jezzard, P.M.
Matthews and S.M. Smith. OUP, 2001.
So I would recommend using this reference.
All GRF methods are only approximate, but despite that, GRF is still
probably the single most widely-used method of inference in SPM and
FSL. GRF is no less "accurate" than when it was first suggested as the
default inference in SPM. In Keith's chapter he refers to Friston94
wrt the first use of spatial extent of cluster; Poline's contribution
at that point was slightly different, looking at multi-scale
statistics, and suggesting cluster-mass, if I remember correctly.
FDR is indeed "newer", but is less widely-accepted as being the
optimal inference approach, not because it is "wrong", but because it
is asking a different question, and many people are not happy with the
question that it asks.
Cheers.
On 27 Nov 2007, at 20:28, Clark Johnson wrote:
> Hi:
>
> We recently got 'dinged' for using the FEAT clusterizing method to
> correct for multiple comparisons. Here is the reviewer's complaint:
>
> "The methods section is at present too confusing to evaluate
> scientifically - just as a side remark - it is not state-of-the-art
> any longer to analyze brain imaging data based on Gaussianized T- or
> F-fields and use corrected P-values based on the cluster-size
> statistic. This was general knowledge at the end of the 90's and I
> doubt Worsley, Evans, Marrett, & Neelin, 1992, even discuss the
> cluster-size statistic, since if I remember correctly, this test-
> statistic was introduced by Poline et al. ~ 1994."
>
> What is your current thinking about this -- should we use the False
> Discovery Rate procedure instead or can we argue that the method we
> are using is both valid and appropriate?
>
> Clark
>
---------------------------------------------------------------------------
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director, Oxford University FMRIB Centre
FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
---------------------------------------------------------------------------
|