Effect size estimation

Discussions
1

Hi!

Is it possible to estimate the effect size (Cohen’s d, rank-biserial coefficient) of the significant values after computing statistical tests (GUI/script)?

Thank you in advance!

2

Hi @DianaCH, for the moment the effect size is not computed when performing tests.

However, Cohen's d is quite simple to compute in Brainstorm.

  1. Compute the difference of means between GroupA and GroupB
    This will result in one file.
    https://neuroimage.usc.edu/brainstorm/Tutorials/Difference#Difference_of_means

  2. Compute the standard deviation for data in GroupA and GroupB
    Use the options Everything and Standard deviation
    This process will result in one file
    https://neuroimage.usc.edu/brainstorm/Tutorials/Averaging#Averaging

  3. Lasly, perform the division of the file from Step1 and the one from Step2

    1. In the Process2 Tab, set the difference-of-means file in FilesA, and the std-of-all-files in FilesB
    2. Run the process: Other > Run Matlab command typing these Matlab commands:
    Data = DataA ./ DataB;

@tmedani, this looks like a useful feature, any thoughts about it?

1 Like
3

Hi Raymundo,

Thank you for the recommendations. Meanwhile, I computed it via script, so I will compare the results from both methods. Any thoughts about how to standardize the W values from the Wilcoxon signed-rank test?

4

Thanks @Raymundo.Cassani for the detailed response.

Yes absolutely, we can add this one and few other used by the community.

@DianaCH Let us know about the comparison of the results.

1 Like
5

yes please! that will really be useful!