Why I am getting nans when calculating the PSD for a sequence of numbers?!!

Hey all,

I have a sequence of data saved in a column vector. I know that Fs = 400 Hz for the data when it was collected in time domain. I want to see the frequency component and calculate the power spectrum density. I used [Pxx,w] = pwelch(x) and all hat i'm getting is Nans in complex and real format!! x is my column vector. Where did I go wrong?!

Answer 1

[Wayne King]

Is there a NaN in your data?

If you have a NaN in your data, that will create the output you describe. For example:

 x = randn(1000,1);
 x(100) = NaN;
 [Pxx,W] = pwelch(x);

You can check for NaNs with isnan()

[Yasmin Tamimi]

Yes, I have nans included in my data! Shall I get rid of them before proceeding with psd estimation??

[Wayne King]

Yes you should, the question is what is a meaningful thing to do because the PSD estimate assumes that the data is uniformly sampled. If you just remove the NaNs that is no longer the case.

[Yasmin Tamimi]

Then I shall linear interpolate the nans and then do the psd estimation.