What it says really! A statistical method that explores the relationship between two variables each observed sequentially.
Details #
This can get fiendishly complicated, unusually, I recommend going to Wikipedia for this only if you have a maths degree! The basic idea is that you suspect that one set of values may predict, to some extent, subsequent values of another. An example might be the heart rate of a client and that of their therapist observed say ever minute across a session.
The cross-lagged correlations in this scenario are the correlation of the client the minute after that of the therapist (therapist to client cross-lagged correlation) and that of the therapist a minute after that of the client (client to therapist). On the face of it these correlations are indicators of what might be a causal effect in one or other direction. There are two issues here: one is that the rate values are not independent of one another, and the other, related, issue is autocorrelation. Autocorrelation is the fact that our heart rates across one minute will be similar to that the minute before and the minute following. To simplify things greatly, there are statistical ways to allow for this autocorrelation and non-independence of observations and get useful indicators of these directional effects, from therapist to client and vice versa. Don’t get into this without a statistician and be wary of reports that are very emphatic about their findings, look for cautious phrasing and listing of the various issues and how they were explored in the particular study. (Of course, this is true when reading any report, see e.g. our BACP paper, originally titled “How to read a paper” but it’s particularly true for cross-lagged correlation and most time series studies.)
Try also #
- Autocorrelation
- Correlation
- Independence of observations
- Null hypothesis significance testing (NHST)
Chapters #
Not covered in the OMbook.
Online resources #
- Our paper for BACP, originally “How to read a paper”.
- I would like to unpack cross-lagged correlation a lot more in the Rblog but when will I get time?!
Dates #
First created 22.i.25.
