forecastCov {dse} | R Documentation |

Calculate the forecast covariance for different models.

```
is.forecastCov(obj)
forecastCov(obj, ..., data=NULL, horizons=1:12, discard.before=NULL,
zero=FALSE, trend=FALSE, estimation.sample= NULL,
compiled=.DSEflags()$COMPILED)
## S3 method for class 'TSmodel'
forecastCov(obj, ..., data=NULL,
horizons=1:12, discard.before=NULL,
zero=FALSE, trend=FALSE, estimation.sample= Tobs(data),
compiled=.DSEflags()$COMPILED)
## S3 method for class 'TSestModel'
forecastCov(obj, ..., data=TSdata(obj),
horizons=1:12, discard.before=NULL, zero=FALSE, trend=FALSE,
estimation.sample= Tobs(TSdata(obj)), compiled=.DSEflags()$COMPILED)
## S3 method for class 'TSdata'
forecastCov(obj, ..., data=NULL,
horizons=1:12, discard.before=1,
zero=FALSE, trend=FALSE, estimation.sample= NULL,
compiled=.DSEflags()$COMPILED)
```

`obj` |
TSdata or one or more TSmodels or TSestModels |

`data` |
an object of class TSdata. |

`discard.before` |
period before which forecasts should be discarded when calculating covariance. |

`horizons` |
horizons for which forecast covariance should be calculated. |

`zero` |
if TRUE the covariance is calculated for a forecast of zero. |

`trend` |
if TRUE the covariance is calculated for a forecast of trend. |

`estimation.sample` |
portion of the sample to use for calculating the trend. |

`compiled` |
a logical indicating if compiled code should be used. (Usually true except for debugging.) |

`...` |
arguments passed to other methods. |

Calculate the forecast cov of obj relative to data. If obj is TSdata then the output data is used as the forecast. For other classes of obj TSmodel(obj) is used with data to produce a forecast. TSmodel() is also applied to each element of ... to extract a model. All models should work with data. If obj is a TSestModel and data is NULL then TSdata(obj) is used as the data. This is multiple applications of forecastCovSingleModel discard.before is an integer indicating the number of points in the beginning of forecasts to discard before calculating covariances. If it is the default, NULL, then the default (minimumStartupLag) will be used for each model and the default (1) will be used for trend and zero. If zero is TRUE then forecastCov is also calculated for a forecast of zero. If trend is TRUE then forecastCov is also calculated for a forecast of a linear trend using data to estimation.sample.

A list with the forecast covariance for supplied models on the given
sample. This is in the element `forecastCov`

of the result. Other
elements contain information in the arguments.

```
data("eg1.DSE.data.diff", package="dse")
model1 <- estVARXar(eg1.DSE.data.diff)
model2 <- estVARXls(eg1.DSE.data.diff)
z <- forecastCov(model1, model2, data=trimNA(eg1.DSE.data.diff))
is.forecastCov(z)
```

[Package *dse* version 2020.2-1 Index]