Hi Steve,
Thank you so much for answering this question. I am looking at data that was processed by someone else and I was concerned about the cluster Z being set to 1. I couldn't understand why the threshold zstat files looked so different from the cluster zstat files. That's why I thought that only the threshold zstat may have been thresholded at the p level, but not the cluster mask.
That would be great if you could take a look. How do I upload the FEAT directory?
Best Regards,
Dana
________________________________________
From: FSL - FMRIB's Software Library [[log in to unmask]] On Behalf Of Steve Smith [[log in to unmask]]
Sent: Wednesday, April 22, 2009 1:04 PM
To: [log in to unmask]
Subject: Re: [FSL] cluster_zstat, thresh_zstat, and no p-values in cluster table
On 21 Apr 2009, at 05:07, Dana Eldreth wrote:
> Dear FSL Experts,
>
> I have several questions regarding the higher level analysis results
> in FSL.
> These questions may be pretty basic, but I've searched everywhere
> and I
> want to make sure that I report results accurately.
>
> A.) I was reading FSL information and I was wondering if the
> "cluster_zstat"
> file shows the activation for clusters that were thresholded at the
> Z=1 without
> considering the thresholded p value. Whereas, the "thresh_zstat"
> file would
> consist of the significant Z=1 clusters that would be thresholded at
> the p=.001
> level.
No - the cluster_zstat text files are all derived from the thresholded
zstat images.
Note that an initial cluster-forming threshold of Z=1 is very low for
inference using random field theory (what FEAT uses at present by
default) - at this low level people seem to think that the theory is
quite inaccurate. We don't generally reduce the initial cluster
threshold below the default of 2.3.
>
> B.) I was looking at the archives to gain further clarification on
> the question
> above and I found this response by Steve Smith by accident. The
> problem
> that we are having is that the clusters coordinates in the cluster
> table all have
> a p-value = 0 and -log10(P) = inf. This is a similar problem to the
> one listed
> below, but there was no follow-up email to this one.
To quote a wise man "this sounds odd - probably worth working out
what's going on".....I've not seen this, even with large
clusters....if you would like to upload the entire FEAT output
directory we can take a look.
Cheers.
>
> Any further clarification is greatly appreciated.
>
> Best Regards,
>
> Dana
> __Email in Archives___
> Hi,
>
> No, this sounds odd - probably worth working out what's going on
> here...do you want to tell us more?
>
> Cheers, Steve.
>
> On 9 Aug 2007, at 11:04, Leon de Bruin wrote:
>> Hi,
>>
>> I'm running cluster-corrected analyses and somehow the cluster-tables
>> display no p-values other than 0, no matter what p-value I used to
>> run the
>> analysis. However, when I run the analysis with a p-value of 1, the
>> clustertable does display significant p-values. Is this an
>> acceptable method?
>>
>> Thanks,
>> Leon
>
---------------------------------------------------------------------------
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
---------------------------------------------------------------------------
|