search

Mensen / ept_TFCE-matlab

Advanced EEG Statistics
27 stars 9 forks source link

2x2 ANOVA #11

Open ArtuDitu opened 6 years ago

ArtuDitu commented 6 years ago

Hey,

great toolbox and PhD thesis.

I have a small problem with understanding what is happening when I use it to do 2x2 repeated measure ANOVA.

The output of the function gives a Results array with 4 cells.

1.Obs These are F-values for comparisons between both factors and interaction for each channel and data point. Please correct me if I'm wrong.

  1. TFCE_Obs These are TFC enhanced F-values from the last step. Here my problem start. Why not enhancing signal first and then calculating F-values?

3.maxTFCE These are max F-values from each permutation of TFCE_Obs.

4.P_Values How are these calculated from maxTFCE and Obs?

I read your thesis and the paper and went through parts of your code. However, I'm left with mentioned above doubts. I would appreciate if you have a second to explain it or point me to the text where I could find the answer.

Thank you in advance.

Mensen commented 6 years ago

Thanks!

  1. Correct. Same values as if you just used a repeated measures ANOVA on a single channel's values.

  2. One could try this and see where the differences are, I'd be interested. However there are a few issues with such an approach I see right away. The first is simply computational: the TFCE procedure is one of the more time consuming in the process and so running it through every participant, for every condition could take substantially longer. The other issue is that the goal is to enhance statistically relevant values. So if certain channels or time points wouldn't be relevant at the level of testing (condition/group level), then it makes little sense to enhance them.

  3. Almost. These are the max TFCE values from each permutation.

  4. These are calculated from the TFCE_Obs and the maxTFCE distribution in a straightforward way. For each point of comparison, what proportion of the maxTFCE values are larger than the observed one.

Hope that helps clarify a few of the points! Let me know if there are things that I couldn't explain clearly enough.

Mensen commented 6 years ago

Hi! The ERPs displayed are the averages for whichever particular factor you are looking at in the viewer. So if you are currently exploring factor A, then the ERPs will be the different levels of factor A. For the individual factors it simply takes the mean over all the levels of the other factors for that level.

The time series is always the original data. The TFCE "correction" is applied only to the T-values of the comparison between factors... so the original data is never altered in any way.

Good luck!