Print

Print
Dear Li,

According to my understanding,

1. Can anyone tell me why the t value is even lower than the uncorrected t value?
    t=2.533749 (p<0.05, FDR corrected) would be higher than the t value of p<0.05 uncorrected. But if you compare p<0.05 (FDR corrected) with p<0.001 uncorrected, then it depends on the data.

2. The difference of FDR correction between SPM and REST .
    SPM's FDR correction is based ONE-TAILED t test. However, it's suggested to based on TWO-TAILED t test for REST's FDR correction in comparing two groups .

3. The difference between SPM and REST's one-tailed FDR correction.
    Basically, the FDR correction in SPM and in REST (if one-tailed selected) are the same. However, I guess you have not choose any mask file in REST FDR correction. In SPM, the FDR correction will use an implicit mask which generated in the "estimate" step (mask.img in the SPM.mat's directory). If you set this mask for REST, then the FDR correction results would be the same.

Hope these information helpful.

Best wishes!

Yours Sincerly,
Chao-Gan


2011/5/6 伪影 <[log in to unmask]>
Hi, everyone
 
     I found a problem. I used SPM5 to process the data and got the signifinance between two groups(p=0.001,t=3.929633, uncorrected). However, When I used FDR method(p=0.05) to correct the result, the output was t=2.533749. I am very confounded. Can anyone tell me why the t value is even lower than the uncorrected t value? In order to verify its reliability, I used another softerware(REST http://restfmri.net/) to test the result with FDR method(p=0.05). Strangely, I found the result was t=3.5232(one tailed) and t=4.2774！Can you explain this? I wonder whether there is some problems with the SPM5 OR REST. Those who offer the help would be greatly appreciated.
 
Li



--
YAN Chao-Gan, Ph.D.
State Key Laboratory of Cognitive Neuroscience and Learning
Beijing Normal University
Beijing, China, 100875
Phone: +86-10-58801023
[log in to unmask]
http://www.restfmri.net/forum/yanchaogan