[R] Test Relation between Time Series? => Cross Correlation Analysis with ccf()

Jan Verbesselt Jan.Verbesselt at agr.kuleuven.ac.be
Mon Mar 15 14:13:24 CET 2004


Dear R world,

When investigating two time series and applying a cross correlation
ccf() function, the results show that correlation is maximal (ccf=0.8)
at a lag of 0.1.

e.g. 
ccf(Inv.KBDIn,ts.medNDII, type=c("correlation"), na.action=na.omit,
main=c("Cross-Correlation of Inverse KBDI against NDII"), ylab="CCF")

1.  Does this mean that there is a positive shift of the first time
series Inv.KBDI against the second?

2.   A lag of 0.1 means that 0.1*Frequency=> amount of time steps. 
e.g.
Frequency=36, => 36 time steps in a year.  So a lag=0.1, multiplied with
36 give us: 3.6 time steps (decades in this case).

*Is this correct?  

*Has anyone other tips for interpretation of (partial)
cross/auto-correlation results? (shape, etc.) (pdf.) or examples of
scripts of how one can extract useful information out of a cross-
correlation plot. Etc. 

*What is the validity of using this as a tool to test the relationship
between to time series vs dependence of results? (max. correlation
coefficient, lag)

Thanks a lot,
Jan


_______________________________________________________________________
Jan Verbesselt 
Research Associate 
Lab of Geomatics and Forest Engineering K.U. Leuven
Vital Decosterstraat 102. B-3000 Leuven Belgium 
Tel:+32-16-329750   Fax: +32-16-329760
http://gloveg.kuleuven.ac.be/




More information about the R-help mailing list