I was working on LFP datasets from some mice. I have a question regarding normalizing the data before time-frequency analysis. Is it wrong to normalize the raw data between 0,1 in microvolt before TF analysis?
We have recorded data from nine mice( 4 normal, 5 Alzheimer ). but I think the distribution of data between mice is different as I don’t get acceptable results while classifying with neural networks.
Does anyone have any suggestions?
Hi Faraz. Normally I would say that this normalization is unnecessary, but many classifiers work better (and/or eliminate biases) when the data are normalized. So yes, min/max scaling seems appropriate to me.
Thanks, Mike. I have another question regarding ITPC. is there any way to create a time-frequency plot that contains both power and phase characteristics of signal? I mean I want to merge them.
Hmm, interesting thought. They are different features of the data, and so normally plotted separately.
I guess you could draw black isocontour lines for the ITPC and plot those lines of top of power? Or use the ITPC to scale the power, such that the resulting plot is literally power.*itpc. That would mean that the power associated with low ITPC would be small. I’m not sure if that’s actually a good idea, just something that comes to mind…
I will check that and get back to you. Thanks Mike.
Hi Mike, Sorry for late update. I tried both methods but there was not any improvemnet in my results. I’m guessing maybe there is a problem with my data too.