I was ready to mention the final point of this lecture, but it was noted. I think scaling is commonly optional in EEG analyses, because at the end of the day the data are compared in terms of relative differences between subjects, condition, etc. Thus, if all the extracted values (that are fed into statistical analyses) are non-scaled, then the final comparisons are fair. Notwithstanding the foregoing, scaling prevents from viewing extreme and "unreal" results :)
Thanks for your comment. Indeed, normalization and scaling are always tricky. It's sometimes necessary, sometimes optional, and sometimes just annoying and distracting.
If the signal is real-valued, then yes, the negative frequencies mirror the positive frequencies (with the exception of 0 Hz, or DC, because +0=-0). Complex-valued signals have asymmetric spectra.
You don't *need* to scale them; it's something you might want to do if you want the coefficients to have the same measurement units as the original data.
I was ready to mention the final point of this lecture, but it was noted. I think scaling is commonly optional in EEG analyses, because at the end of the day the data are compared in terms of relative differences between subjects, condition, etc. Thus, if all the extracted values (that are fed into statistical analyses) are non-scaled, then the final comparisons are fair. Notwithstanding the foregoing, scaling prevents from viewing extreme and "unreal" results :)
Thanks for your comment. Indeed, normalization and scaling are always tricky. It's sometimes necessary, sometimes optional, and sometimes just annoying and distracting.
Are the negative frequencies always symmetrical with the positive frequencies?
If the signal is real-valued, then yes, the negative frequencies mirror the positive frequencies (with the exception of 0 Hz, or DC, because +0=-0). Complex-valued signals have asymmetric spectra.
First slide ... Did you mean to write [Normalizing]?
[Normalize] ?
i don't underand why we need to scale fourier cofficents
You don't *need* to scale them; it's something you might want to do if you want the coefficients to have the same measurement units as the original data.