clustering after alphasim

Submitted by zswgzx on Wed, 01/07/2015 - 18:06

Hi DPABI experts,

I'd like to perform a statistical analysis, i.e. two-group ttest using the z-transformed functional connectivity maps, and have already done the model estimation.

Now I'm trying to do multiple comparison correction with alphasim and have questions about the procedures.

So I open Viewer from dpabi main menu, import the overlay T2.nii from model estimation, and choose AlphaSim correction for computation. How should I use the output AlphaSim.txt for extraction of significant clusters afterwards?

Also the min. cluster size with an alpha <0.05 is 983. Does it make sense? I ran 10000 iterations with a brain mask.

Finally, how can I view the clustered results and extract statistics such as cluster size, max statistical intensity in the cluster, etc.?

I've viewed the course ppt but still don't know how to do the above.

Any help/suggestion is appreciated.

Shengwei

Forums:

YAN Chao-Gan

Thu, 01/08/2015 - 19:44

Hi Shengwei,

Once you got the min cluster size with alpha < 0.05, you can set it in "Set Cluster Size" under "cluster" menu.

Here you got 983, it's quite big. What's the smoothness? Also, have you tried GRF corredtion as well, what's the cluster size there?

You can click "Cluster Report" under "Cluster" menu to get the relevant report.

Best,

Chao-Gan

zswgzx

Tue, 01/13/2015 - 17:50

Thank you for the instructions, Chaogan. Smoothness is about 13mm, and similar cluster size was calculated for GRF correction.

Shengwei



YAN Chao-Gan

Wed, 01/14/2015 - 16:41

Hi Shengwei,

The large cluster size is because you have large smoothness. Try not to put big smooth kernel in preprocessing.

Best,

Chao-Gan

YAN Chao-Gan

Wed, 01/14/2015 - 16:41

Hi Shengwei,

The large cluster size is because you have large smoothness. Try not to put big smooth kernel in preprocessing.

Best,

Chao-Gan

zswgzx

Fri, 01/30/2015 - 18:29

Thanks, Chaogan. I assume the "smoothness" you mentioned is based on the statistical image with t values, not the one used in the preprocessing after normalization to a template. Is it right?
I have one more question related to alphasim: how should I choose the appropriate voxel p value?

YAN Chao-Gan

Wed, 02/04/2015 - 21:33

Hi,

Right, "smoothness" is not the smooth kernel used in preprocessing. It can be estimated from the 4D residuals of Statistical Analysis (DPABI will do that automatically).

For voxel p value, either 0.05 or 0.01 is common in literature, and should be acceptable.

Best,

Chao-Gan