why are values returned by xcorr not within [-1,1]?

I am using [r, lags] = xcorr(y1,y2) where y1 and y2 are two time series data.

I would like to know the meaning of the values returned by r and why they are the correlation values I know that should be in [-1,1].

Your insights would be great.

Answer 1

[Adam]

Use the 'scaleopt' option as documented in the help if you want normalised results. Otherwise the results will depend on the magnitude of your data.

e.g.

a = 1:10;
b = 1:10;

xcorr( a, b )

will return values ranging from 10 to 385.

xcorr( a, b, 'coeff' )

will return values between 0 and 1;

[Rafael Miranda]

thanks for your contribution!

Could you explain what this 'coeff' means?

[Adam]

doc xcorr

explains it.

[michal.markun]

I don't understand it. Correlation should always be higher than -1 and lower than +1, by definition, for any lead/lag and and irrespective of whether you compute it for the same variable (auto-) or for two variables (cross-).

[Torsten]

@michal.markun

Maybe you think of "corr":

https://uk.mathworks.com/help/stats/corr.html