Y offset changes autocorrelation
kgerhold
I correlate a data set with itself (autocorrelation/serial correlation) to see whether or not it has an oscillatory pattern. If the Y values are negative, Igor seems to have problems finding any correlation, although it looks oscillating. Additionally a Y offset will change the shape of the correlation.
The attached image shows the same graph with different Y offsets (left) and the respective autocorrelation (right).
Could anyone explain to me, why there is a difference in the autocorrelation depending on the Y values of the data? Is the correlation of data sets with negative Y values reliable?
Any help or suggestions are greatly appreciated.
Consider your offset as a rectangular function with a width identical to your data x range. Autocorrelation of a rectangular function gives a triangular shaped result. This is what you see in your example (note that you multiply two negative values in case of y<0 and the result is still positive).
You probably already had a look, but anyway:
https://en.wikipedia.org/wiki/Autocorrelation
http://pages.jh.edu/~signals/convolve/index.html (Hope it still works. Very illustrative.)
HJ
July 28, 2015 at 05:44 am - Permalink