autocorrelation for circular variables?

[Mark Zarella]

Suppose I have angular data. How do I perform an autocorrelation on this
data taking into account the fact that 359 degrees, for example, is more
closely correlated to 0 degrees than would be revealed in a typical linear
correlation function? Would it be valid to replace products with dot
products in the integral? Any ideas or references would be greatly
appreciated.

[Colin Andrew Percival]

Post by Mark Zarella
Suppose I have angular data. How do I perform an autocorrelation on this
data taking into account the fact that 359 degrees, for example, is more
closely correlated to 0 degrees than would be revealed in a typical linear
correlation function?

Map x -> exp(i x) before computing the autocorrelation.

Colin Percival