Abstract
Many time series can be observed at different, but equally relevant sampling frequencies.
This makes it important to study aggregation from e.g. daily or weekly to monthly series.
Aggregation of course gives shorter time series and thereby reduced information, but spurious
phenomena, in e.g. daily observations, can on the other hand be avoided such that more important
features become clearer. In the present study we contribute to the literature on temporal
aggregation of cointegrated time series by giving a theorem for how the speed-of-adjustment
coefficients in a n-dimensional VAR(1) process changes with the frequency of the data. We also
introduce a graphical representation that will prove useful as an additional informational tool for
deciding the appropriate cointegration rank of a model. In two examples based on models of time
series of different grades of gasoline, we demonstrate the usefulness of our results in practice.
This makes it important to study aggregation from e.g. daily or weekly to monthly series.
Aggregation of course gives shorter time series and thereby reduced information, but spurious
phenomena, in e.g. daily observations, can on the other hand be avoided such that more important
features become clearer. In the present study we contribute to the literature on temporal
aggregation of cointegrated time series by giving a theorem for how the speed-of-adjustment
coefficients in a n-dimensional VAR(1) process changes with the frequency of the data. We also
introduce a graphical representation that will prove useful as an additional informational tool for
deciding the appropriate cointegration rank of a model. In two examples based on models of time
series of different grades of gasoline, we demonstrate the usefulness of our results in practice.
| Original language | English |
|---|---|
| Journal | Advances and Applications in Statistical Sciences |
| Volume | 6 |
| Issue number | 4 |
| Pages (from-to) | 207-227 |
| Number of pages | 28 |
| Publication status | Published - 2011 |