Print

Print
Hi Donald,

If I use an uncorrected p value of 0.01, then when printing the tableI have corrected clusters .

Does this mean the same that you said.

Regards,

AS

On 24 Oct 2013, at 08:41 am, "MCLAREN, Donald" <[log in to unmask]> wrote:

If the corrected p-value is 0.01, then it means that that cluster would survive the cluster threshold of 0.01 or less and 0.05 or less (as 0.05 is less significant than 0.01).

Best Regards, Donald McLaren
=================
D.G. McLaren, Ph.D.
Research Fellow, Department of Neurology, Massachusetts General Hospital and
Harvard Medical School
Postdoctoral Research Fellow, GRECC, Bedford VA
Website: http://www.martinos.org/~mclaren
Office: (773) 406-2464
=====================
This e-mail contains CONFIDENTIAL INFORMATION which may contain PROTECTED
HEALTHCARE INFORMATION and may also be LEGALLY PRIVILEGED and which is
intended only for the use of the individual or entity named above. If the
reader of the e-mail is not the intended recipient or the employee or agent
responsible for delivering it to the intended recipient, you are hereby
notified that you are in possession of confidential and privileged
information. Any unauthorized use, disclosure, copying or the taking of any
action in reliance on the contents of this information is strictly
prohibited and may be unlawful. If you have received this e-mail
unintentionally, please immediately notify the sender via telephone at (773)
406-2464 or email.


On Thu, Oct 24, 2013 at 12:33 AM, fMRI <[log in to unmask]> wrote:
Dear Christophe,

A very basic follow up question, if I use a p value at the voxel level of  say 0.001, in the SPM result I get something at the cluster level or coxel level that says corrp, which refers to the corrected p value. 

My question is, using this p value 0.01, does this mean that this cluster is corrected at p value of 0.01 or 0.05? 

Regards,

AS

On 14 Oct 2013, at 09:09 pm, Christophe Phillips <[log in to unmask]> wrote:


Hi Aser,

Cluster-level inference works in 2 steps.
First you need to define the clusters. This is done by applying a threshold at voxel level to form clusters, i.e. sets of spatially connected voxel above this threshold. Using .001 or .0001 will indeed give you a different set of clusters: larger with .001 than .0001obviously.
Second you infer the probability of observing by chance (null hypothesis) a cluster of that size given the smoothness in your data. Large clusters are unlikely to appear 'by chance' and therefore receive a small p-value. You get a corrected p-value by accounting for the fact that you're searching through a volume and thus could get "many" clusters by chance.

Note that with cluster level inference you're more sensitive, i.e. it's "easier" to get significant results, but less specific, i.e. your inference is about the whole cluster and so you cannot report any specific location.

HTH,
Chris


-

De: "fMRI" <[log in to unmask]>
À: [log in to unmask]
Envoyé: Lundi 14 Octobre 2013 21:47:08
Objet: Re: [SPM] Showing just significant clusters

Hi

Thanks for your response

I am not sure if I understood what you meant.

Basically when I open any result. If I select immediately FWE, I get very few voxels. If I change this to non corrected 0.001 or 0.0001 and then press whole brain result  and then take the min size of the FWEc and then press again 0.0001 and enter that min FWEc as extent threshold, I get all the significant clusters at corrected values. 

My question is when I find the min FWEc using a threshold of 0.0001 uncorrected and take the min size and enter it again using 0.0001, what should I understand from this step? Is that I use a height threshold of 0.0001 to identify corrected clusters at 0.05 or at 0.0001? It might be a naive question but confusing for me.

  

Regards,

Aser

On 14 Oct 2013, at 06:50 pm, "Canterberry, Melanie" <[log in to unmask]> wrote:

Hi –you don’t need to change the p-value. Just change the minimum number of voxels that you enter. When you look at the original stats table, you should find the smallest cluster size that meets .05/FWE. Then just re-do the analysis choosing the same p-value as your original analysis, but change the number of voxels to the number your found from the stats table. This will keep everything the same, only you will only see the clusters that meet FWE. You won’t need to change your description as you will be showing the clusters that meet your originally defined voxel significance criteria at the cluster level.

 

Hope that helps.

 

Melanie

 

From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]] On Behalf Of Aser A
Sent: Monday, October 14, 2013 10:16 AM
To: [log in to unmask]
Subject: Re: [SPM] Showing just significant clusters

 

Hi Guillaume and all,

If I select significant clusters by first seeing the the minimum k and then re-defined the the threshold to be ( 0.0001). Then How can I describe this? ...Significant corrected (FWE) clusters selected using a threshold of 0.0001 ?! Because selecting 0.001 or 0.0001 will split the clusters and change its numbers ...etc

 

On Mon, Sep 23, 2013 at 1:28 PM, Guillaume Flandin <[log in to unmask]> wrote:

Hi Aser,

to do so, you can figure out from the results table what the cluster
size critical threshold is (the minimum of the column k that has a
cluster-level corrected p-value lower than 0.05). This is also reported
in the table footnote on the bottom left (FWEc).
Then click on Results again and this time enter this value when prompted
for 'extent threshold'.
This is illustrated at some point in the middle of this webpage:
http://imaging.mrc-cbu.cam.ac.uk/meg/SensorSpm

Best regards,
Guillaume.



On 21/09/13 14:26, Aser A wrote:
> Hi Donald and all,
>
> I would like to view just significant clusters ( corrected FWE). Is
> there a way to do this ? For example using Five. If yes how ?
>
> Thanks
>
> Aser

--
Guillaume Flandin, PhD
Wellcome Trust Centre for Neuroimaging
University College London
12 Queen Square
London WC1N 3BG