I calculated an Spectrogram of invasive electrophysioloy of not long time series. We could not see any evident clue in the "heatmap" of Spectrogram, besides the power of some frequency bands seems to oscilate along time. So, now we are trying to look at each frequency as a time series of Power Spectral Density (PSD), by using the PSD along time for a particular frequency and estimate the PSD of it. I already tried to apply the Welch's and Multitaper methods, even the Periodogram (through Matlab functions), but I got sort of random peaks, without any evident clue again. For the very first Spectrogram I had 250 Hz sampling using sliding temporal window, but for the "second" estimation of PSD got only 43 Hz, because of the sliding temporal window.

Could you please give some advice or suggestion?