ITPC, multiple comparisons and random distribution.

[Casper Kerren]

Dear Mike,

I have a fourier output of my data, and want to investigate the ITPC at a certain time point and frequencies. 

cfg               = [];

cfg.keeptrials  = 'yes';                       

cfg.method      = 'wavelet';                  

cfg.width       = 6;                           

cfg.output      = 'fourier';                 

cfg.foi         = 1 : 1 : 20;                   

cfg.toi         = -2 : 0.05 : 2;               

freq            = ft_freqanalysis(cfg, data);

This seems pretty straightforward, but the problem is that I only have one condition. I can create a null distribution, but the way my pipeline is built I am afraid I would have to go through the entire preprocessing stage again. Something I rather not do (of course, if this is the best way to do it, I will). I have used Rayleigh's Z (page. 488, equation 34.3) to compute my p-values. I have also used equation 34.4 to find the critical value of my ITPCs. Perhaps my interpretation of the Rayleigh's Z or the critical values is wrong, but the Rayleigh's seems to give me statistical significant results, whereas the equation to find the critical values contradicts this. 

I do have a lot of data points and as you also mention: they are not well suited for controlling for multiple comparisons. 

I am not sure I have made my point, but my main questions are: Can I use Rayleigh's Z or the equation to find the critical value when I have a huge amount of data points, and in that case, how do I control for multiple comparisons? Second: Is the best alternative to create a random distribution and use that as my second condition? 

Kind regards,

Casper

[Mike X Cohen]

Dear Casper,

If it's not convenient or possible to use permutation testing to find a null hypothesis of cluster sizes for multiple comparisons corrections, you could also try another approach such as FDR for example.

I'm not sure what you mean by 34.3 and 34.4 contradicting each other. They are inverses of each other (adding the negative sign to prevent taking the square root of a negative number), and so should be comparable.

Mike

--
You received this message because you are subscribed to the Google Groups "AnalyzingNeuralTimeSeriesData" group.
To unsubscribe from this group and stop receiving emails from it, send an email to analyzingneuraltimeseriesdata+unsub...@googlegroups.com.
Visit this group at https://groups.google.com/group/analyzingneuraltimeseriesdata.
For more options, visit https://groups.google.com/d/optout.

--

Mike X Cohen, PhD
mikexcohen.com

[Casper Kerren]

Thank you for your fast reply. 

Contradict is not the correct word to use, sorry for that. I mean that my interpretation of the results from the two equations contradict each other. Probably something wrong in how I implement it in Matlab. 

Will look into the FDR, thanks!

Casper