Closed rkillick closed 2 years ago
Yes, auto-covariances are probably more common, but there may me non-auto-covariances in models with multiple variables, and I would think that these standards could apply as well. So perhaps no need for 'auto'? No strong opinion, though.
I would argue that covariance would come under standards for multivariate data, and that multivariate time series would have to adhere to those as well as the time series standards. Autocovariance is something more specific to time series that isn't seen as much in other areas (unlike covariance).
Thanks @rkillick and @christophsax for clarifying and documenting this discussion for future reference. I'm happy to merge now - covariance calculations are (almost always) general and irrespective of whether or not points are ordered by time or along some other, atemporal dimension, so exactly as Rebecca suggested, it is more accurate to refer to the matrices here as "autocovariance" structures.
Grammatical corrections.