Dear Mr. Espinos,
SPM presents significance levels for both clusters and peak-level activation. In both cases you are given an uncorrected P value, an FWE corrected P value and a Q value. This Q value is obtained using an FDR-correction. Regardless of what significance threshold you choose you will be given all 3 values for both cluster and peak values.
P values of 0.000 simply indicate that the value is lower than 0.0004, yet SPM displays just 3 decimals.
The reason you find different statistics for your clusters is because dropping the significance level will change the size of the cluster, which in turn changes the calculation for this cluster.
I hope this answers your question.
On Fri, 12 May 2017 12:56:03 +0200, Hector Espinos <[log in to unmask]> wrote:
>Good afternoon SPM Users!
>I have a little problem with the SPM results interpretation. I refer to
>the Statistics results.
>I put p<0.001 uncorrelated. The High Threshold F have a determined
>value. I understand that this value is the value with that is
>acomplished the p-value<0.001.
>
>The statistics SPM report have three parts (set level, cluster level and
>peak level.
>
>In the case of cluster level what is the meaning of PFWE-corr, QFDR-corr
>y Punc. I understand that PWE-corr would be the value of the p-value if
>we apllied a FWE correction, the same with the QFDR, in this case with
>False Detection Rate, but I don't understand well the meaning of the
>p-value uncorrelated. I believe that is the p-value in the cluster. But
>what is the meaning when this value is 0.000?.
>
>I probe for example with p<0.02 (i increase the signal) but in theory i
>have less accuracy in my results, but in this case the pfwe and qfdr and
>punc are better. What is the meaning of this?
>
>Thanks in advance!
|