1
$\begingroup$

I want to know that the addition of a uncorrelated time series x with a correlated time series, say y generates a time series but is that correlated or uncorrelated output? For example: x is White Gaussian noise and y is pink noise, then z = x+y Looking at the plot which is the output of the noisy time series, it is hard to say whether z is correlated or uncorrelated. Is there a rule about the nature of the output?

enter image description here

$\endgroup$
2
  • $\begingroup$ Can you re-consider your question a little? You begin with x being a correlated time series and y an uncorrelated series, but in the very next sentence, x is white Gaussian noise (which most people would regard as an uncorrelated series) and y is pink noise (which most people would regard as a correlated series). So, which is which? and what, if any, is the cross-correlation between x and y? $\endgroup$ – Dilip Sarwate Sep 1 '20 at 2:03
  • $\begingroup$ @DilipSarwate: sorry for the typo. I have fixed it. $\endgroup$ – Sm1 Sep 1 '20 at 2:17
2
$\begingroup$

Are $x$ and $y$ uncorrelated? In this case the autocorrelation function of $x+y$ is just the sum of the autocorrelation functions of $x$ and $y$, which if one of them is not white will lead to its sum also not being white.

In general, for stationary processes you have $\phi_{zz}(\tau) = \phi_{xx}(\tau) + \phi_{yy}(\tau) + \phi_{xy}(\tau) + \phi_{yx}(\tau)$ where $\phi_{xy}(\tau) = \mathbb{E}\{x(t)y(t+\tau)\}$. Moreover, since the process is stationary you also have $\phi_{xy}(\tau) = \phi_{xy}(-\tau) = \phi_{yx}(\tau)$. So in general, the autocorrelation of $z$ depends not only on the autocorrelations of $x$ and $y$ but also their cross correlations.

$\endgroup$
4
  • $\begingroup$ thank you for your answer, but it does not address completely my problems. One variable has a flat power spectrum and so is white (x r.v ) and the other has a non flat spectrum so is colored, consider say brown color (y r.v). Since x is white it is uncorrelated to itself and y is correlated to itself (each random number for the red noise is positively correlated to the one preceding it). In general, does the addition of an uncorrelated r.v with a correlated r.v give an output that is correlated or uncorrlelated? i.e., z=x+y is z correlated/uncorrelated? $\endgroup$ – Sm1 Sep 1 '20 at 12:19
  • 1
    $\begingroup$ It does answer your question. If your variables are independent, the correlation of the sum is equal to the sum of the correlations. Therefore, if one is not white (your $y$), the sum $x+y$ is also not white (i.e., samples are correlated to each other correlated), as I wrote in my second sentence. $\endgroup$ – Florian Sep 1 '20 at 13:59
  • $\begingroup$ Thank you, now it is clear. $\endgroup$ – Sm1 Sep 1 '20 at 15:58
  • $\begingroup$ You're welcome :) Glad it helped. $\endgroup$ – Florian Sep 1 '20 at 16:45

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.