topic Re: Time series cross correlation with lag in SAS Data Science

Time series cross correlation with lag

Peder — Mon, 21 Oct 2013 19:53:28 GMT

Hello, I want to check if some temperature data that I have are correlated, and I suspect there is a time-dependent lag.

I have tried the time-series analysis using cross correlation, but either I don't understand how to do it or something is wrong.

Finally I tested with the following data columns, where there should be full correlation for a lag of 7 rows:

A A+7

1 8

2 9

3 10

4 11

etc.

What I got just doesn't make sense. There should be some way to see a full correlation of the data for a lag of 7.

jwexler — Wed, 23 Oct 2013 13:24:34 GMT

Hi, I received a response from one of my R&D colleagues:

EM 13.1 includes the cross-correlation analysis in TS Correlation node.

However, EM users can do cross correlation analysis using proc timeseries if they have ETS license.

The following pseudo code will calculate cross-correlation between t1 and t2.

proc timeseries data=yourdata outcrosscorr=_out ;

id MNTH_YR interval=MONTH accumulate=TOTAL;

var t1;

crosscorr CCF CCFNORM lag n / nlag=24 ;

crossvar t2;

run;

Peder — Fri, 25 Oct 2013 14:53:35 GMT

Thank you.

I am looking for an answer in SAS JMP.

Unless I am doing somewhing wrong the Time Series lag correlation is not working.

I have tested Minitab, where it is working as it should.

jwexler — Fri, 25 Oct 2013 17:33:23 GMT

If that doesn't work, try filing a Tech Support track at support.sas.com. They can look at your code and walk you through any remediation steps.

Thanks,

Jonathan