The problem with this method is that, unless I'm mistaken, all this means is that your peak voxel survives the corrected threshold, while all you know about the other voxels is that they survived the initial uncorrected threshold, which in the case of 0.01 or 0.05 would be quite liberal. This is because when using the SVC button in SPM, there isn't another thresholding step being done, rather it is just computing the corrected p-values for your existing voxels based on the reduced search volume.
I think a better way to do something like this would be to use a mask (your ROI(s)) on your results and then specify a corrected voxelwise threshold. This ensures that all the displayed voxels survive the threshold, but the correction will not be as strict as whole brain since you are pre-specifying a search region.
-Mike
> -----Original Message-----
> From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]]
> On Behalf Of Mike
> Sent: Wednesday, August 31, 2016 1:14 PM
> To: [log in to unmask]
> Subject: [SPM] initial threshold for small volume correction
>
> Hi,
>
> Perhaps a basic question but I can not find a conclusion from previous posts.
>
> While performing svc, the step is usually as follows:
>
> (1). an intial uncorrected threshold, e.g. 0.001 (2). select "small volume" in
> SPM8 (3). define the small volume, either a sphere centered at predefined
> coordinates (e.g. from previous literature) or a mask (e.g. from an anatomical
> mask in an atlas) (4). check the peak-level threshold. If a voxel survives FWE
> 0.05, it is regard as significant.
>
> My question is for (1). It seems 0.001 and 0.005 uncorrected are popular, but
> can I use 0.01 uncorrected, or even 0.05 uncorrected? (btw, I found many
> articles did not describe the initial threshold they used in svc)
>
> Thanks. Mike
**********************************************************
Electronic Mail is not secure, may not be read every day, and should not be used for urgent or sensitive issues
|