Forecasting Multiple Days Ahead with an Asymmetric DCC-GARCH Model in R

Summary

The exchange discusses whether an asymmetric dynamic conditional correlation GARCH model can produce covariance forecasts beyond one day using R’s rmgarch package. The questioner believes the package supports only one-day forecasts and asks how to obtain an n-day forecast. The replies say that multi-day forecasts are possible and recommend setting the forecast horizon with the package’s n.ahead argument in dccforecast.

One response gives a two-day example, while another reports having used a five-day horizon and emphasizes the argument name. The exchange therefore offers a concise implementation pointer rather than a derivation of the asymmetric DCC-GARCH forecasting equations or a validation of forecast quality. It provides no sample data, code context, or discussion of how the package handles future conditional correlations and covariance aggregation. Users should treat it as guidance on specifying the requested horizon and consult the package’s documentation when adapting the call to a particular fitted model or version.

Key ideas

• The replies state that rmgarch can generate asymmetric DCC-GARCH forecasts over multiple days.
• The forecast horizon is specified with the n.ahead argument in dccforecast.
• The exchange includes examples of two-day and five-day forecast horizons.
• The responses provide an implementation pointer but do not explain the forecasting mathematics or assess forecast accuracy.

Tags

# $n$-day ahead forecast for asymmetric DCC-GARCH model


# $n$-day ahead forecast for asymmetric DCC-GARCH model












I am working on forecasting covariances with the use of MGARCH models. I was wondering if anyone knows how to implement a n-day ahead forecast of the aDCC (asymmetric DCC) model in R. The `rmgarch` package in R only supports 1-day ahead forecasts

## Answer by Alba (score 0)

https://quant.stackexchange.com/a/54444

I was under the impression that it is possible in the rmgarch package. For 2 days, you could try:

```
dccforecast(your_model_fit, n.ahead = 2, ...)
```

## Answer by Guido (score 0)

https://quant.stackexchange.com/a/70220

I recently did that for 5 days, so it possible to do it. Just make sure to use n.ahead and not n.head. If you have further errors just post the code here. Anyway Alba already suggested the correct code to use.

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.