Abstrakt
The analysis of time varying conditional correlation structures seems to be a significantly important part of multivariate time series modelling, particularly from the (practical) financial or economic point of view. In~2002, Robert Engle published an innovative concept in the framework of this issue.
A simple class of multivariate autoregressive conditional heteroskedasticity models, the so-called dynamic conditional correlation models were introduced. Thereafter, these techniques have been examined and adjusted in many different theoretical or empirical ways.
In the contribution, several various approaches to modelling the dynamic conditional correlations originally based on Engle's idea are reviewed and discussed. Some of their pros and cons are mentioned and demonstrated.
Finally, the comparison of their performance is shown in the study of the portfolio of the European currencies and their correlation links. All the relevant procedures are implemented in the statistical software R.