It seems like the PACF graph is more informative than an ACF graph. What information does one get from an ACF graph that can not be gleaned from a PACF graph. (ACF = Auto-correlation function, and PACF = partial Autocorrelation).
1
-
2$\begingroup$ I was taught that one can potentially identify the difference between an auto-regressive series and a moving average series by examining the PACF (see this webpage by Robert Nau for an explanation). A more complete answer would state why that is the case though (and functionally what is the difference between an AR and a MA process mathematically). $\endgroup$– Andy WCommented Nov 22, 2011 at 13:48
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
2
They are algebraically relatable http://en.wikipedia.org/wiki/Autoregressive_model#Yule-Walker_equations thus there is no more information in one than the other. What do they speak is the unconditional relationship ACF and the conditional relationship PACF. If one has the ACF then one can impute the PACF and vice-versa.
-
$\begingroup$ This assumes the process is AR, right? Whereas the ACF and PACF are tools that work with any stationary series, whether it's AR or not. $\endgroup$– whuber ♦Commented Nov 22, 2011 at 15:38
-
$\begingroup$ :whuber I don't think so . I think in general there is a relationship between the two acf's although I could be wrong ! $\endgroup$ Commented Nov 22, 2011 at 22:43