Dear Experts,
I read in a previous post called "ICA melodic eigen spectrum" about some strange drop in the eigenspectrum graph as being reported by melodic. It is suggested to have a look at the single session ICA (but what exactly should I attend to?).
Now, when looking at my data, I also see a rapid drop in the eigenvalues (see attachment EVplot2) even stronger than in the previously mentioned post. However, when running the analysis with a smaller subgroup of the original one I got a nice asymptotic drop in the eigenvalues (attachment EVplot). I don't know what this exactly tells about my data. Did something went seriously wrong or is it just due to my data? (1) So what causes this dramatic drop in the eigenspectrum (is it due to artifacts, outlier, missing data, noise, ...?) and (2) what implications for my analysis does this have? Should I be concerned about the interpretation of my results? (3) Is there a possibility to counter this drop?
I thank you for your help!
With kind regards,
Stefan Lenzen
|