[R] cross correlation of filtered Time series

Sudheer Joseph sjo at incois.gov.in
Mon Jan 11 05:58:17 CET 2016


Dear Experts,

I have been trying to find lagged correlation between two time series  which are band pass filtered  for 30-90 day band. I would like to understand the lagged correlation between these 2 series in this band.

The issue is that auto-correlations exist in both series and( also both have the effect of filtering if any).  I am unclear on how to separate out these factors and find the true lagged relation if any among the 2 series. Apparently there is a strong lagged correlation between the two series around 10 days lag. And there is strong ACF in both series at around 20 days. So it is difficult for me to interpret the cross correlation.

If your time permits can you please help me with your advice on this?

The plot is at below link:

https://drive.google.com/file/d/0B3heUQNme7G5UEpzWjd2WmxYdVk/view?usp=sharing

The data is at below link:

https://drive.google.com/file/d/0B3heUQNme7G5T2JjeTlDNzdkX00/view?usp=sharing

R code:

dc = read.csv('test_dat.csv', header=TRUE)
par(mfrow=c(3,1))
acf(dc$sst, 36)
acf(dc$t2m, 36)
ccf(dc$sst, dc$t2m, 36)

with best regards,
Sudheer


More information about the R-help mailing list