This is a general statistics question. Let's say I have a set consisting of several time series and I would like to find pairs that have correlation. Is the only way to do that is to compare each pair one by one? It does not seem so convenient since set may be very large.
Asked
Active
Viewed 17 times
1 Answers
0
It is the easiest and it is broadly implemented in software packages. This help to identify others variables linked.
For time series is very common to do it also with lagged and forward variables. This help to identify links in terms of the future and the past and the same variable and/or others.
Rafael Valero
- 2,736
- 18
- 28
-
So you mean that I have to compare each pair one by one? Time complexity of this is O(n^2) which is not so desirable. – Harun Türkmen Feb 17 '21 at 08:36
-
Yes, but doable. If you 1000 variables it is perfectly possible. Also it is something easy to parallelize. – Rafael Valero Feb 17 '21 at 08:40