Hi,
The first command applies a mask to your statistical map. This can be
the whole brain mask, or an ROI that you specify. It also applies 2
thresholds: a Z threshold that determines which voxels are considered
significant, and a cluster extent threshold (the p value in the GUI),
which determines the minimum size a cluster must have to be considered
significant. You must use this threshold to determine the minimum size
for a cluster.
The second command is run without applying a mask, and does not specify
any cluster extent threshold, which I assume means that it runs
voxelwise on the whole brain.
I don't think you can specify directly how many voxels a cluster should
be to be considered significant. Rather, you use the cluster extent
threshold (p threshold) to specify a maximum probability of type I error
that you consider acceptable. For example, setting p<0.05 means that
there is a probability < 5% that the clusters appearing in your
resulting stat map are due to chance. This takes in account the
properties of your data, i.e. spatial resolution, smoothness, etc..., to
determine the minimum size a cluster should be to meet these criteria.
There has been several more detailed discussions about how thresholding
works - have a look in the archives.
And of course, if I'm forgetting anything or getting something wrong,
please someone correct me!
Best,
Stephane
Brittany Copp wrote:
> I am still a bit confused about the difference between the two options for
> easythresh. How do these two options differ when interpreting data?
>
> Usage: easythresh <raw_zstat> <brain_mask> <cluster_z_thresh>
> <cluster_prob_thresh> <background_image> <output_root> [--mm]
> e.g.: easythresh stats/zstat1 mask 2.3 0.01 example_func grot
>
> Or: easythresh <stat> <stat_thresh> <background_image> <output_root>
> e.g.: easythresh stats/zstat1 2.3 example_func grot
>
> Is there anywhere to specify how many voxels constitute a cluster?
>
> Thanks!
>
> Brittany
>
>
>
|