# Thread: Can some one help explain cross correlation analysis?

1. ## Can some one help explain cross correlation analysis?

Hello,
I was wondering if someone could help explain cross correlation to me with the data i have available.
I have two time series
One is how variable X changes each month over 133 months
the other is how variable Y changes month over 133 months

I'm using mini tab and i've set the LAG to 25% (33) as i wasn't sure what it should be and it produces a graph that i've copied in.
Can someone give me an insight into what i'm looking at?
I know this sounds really stupid but i'd love for someone to explain what the chart is actually demonstrating.

2. ## Re: Can some one help explain cross correlation analysis?

Originally Posted by joshnathan
Hello,
I was wondering if someone could help explain cross correlation to me with the data i have available.
I have two time series
One is how variable X changes each month over 133 months
the other is how variable Y changes month over 133 months

I'm using mini tab and i've set the LAG to 25% (33) as i wasn't sure what it should be and it produces a graph that i've copied in.
Can someone give me an insight into what i'm looking at?
I know this sounds really stupid but i'd love for someone to explain what the chart is actually demonstrating.

Hi joshnathan!

The cross correlation indicates how much X and Y are related and in particular how much variable Y lags with respect to X.
In the diagram you see that there is a peak is about 5 months.
This means that Y is similar to X, but lags about 5 months behind X.
This is not very clear though, because we see a high correlation between 0 up to 33 months.
However, we see specifically that Y is not really ahead of X, since the cross correlation is low there.

The LAG of 33 that you specified limits the graph to 33 months behind up to 33 month ahead.

3. ## Re: Can some one help explain cross correlation analysis?

Hello thanks with the useful information,
would it be worthwhile for me to produce one with a lag of 133 to see if it changes at any point?
This way i can analyse if there is a greater lag in the future.
Based on what you've said there is very little i can draw from this really as the correlation is relatively high for a very large period.

One thing i also wanted to confirm was that cross correlation treats both variables as being in a time series correct as these 133 events for x and y occur 1 month after each other? sorry if it seems a bit stupid just i know that normal correlation doesn't observe the time series nature of the data.

4. ## Re: Can some one help explain cross correlation analysis?

Originally Posted by joshnathan
Hello thanks with the useful information,
would it be worthwhile for me to produce one with a lag of 133 to see if it changes at any point?
This way i can analyse if there is a greater lag in the future.
Based on what you've said there is very little i can draw from this really as the correlation is relatively high for a very large period.

One thing i also wanted to confirm was that cross correlation treats both variables as being in a time series correct as these 133 events for x and y occur 1 month after each other? sorry if it seems a bit stupid just i know that normal correlation doesn't observe the time series nature of the data.
Yes.
A longer lag than say 25% does not necessarily give more information though.
If the overlap in time of the series becomes too low, the numbers won't mean much.