tsfilter.tsmarch function

Model Filtering

Filters new data based on an already estimated model.

## S3 method for class 'cgarch.estimate'
tsfilter(
  object,
  y = NULL,
  newxreg = NULL,
  update = TRUE,
  cond_mean = NULL,
  ...
)

## S3 method for class 'dcc.estimate'
tsfilter(
  object,
  y = NULL,
  newxreg = NULL,
  update = TRUE,
  cond_mean = NULL,
  ...
)

## S3 method for class 'gogarch.estimate'
tsfilter(object, y = NULL, newxreg = NULL, cond_mean = NULL, ...)

Arguments

• object: an object of class cgarch.estimate or dcc.estimate .
• y: an xts matrix of new values to filter.
• newxreg: not used in these models.
• update: whether to update certain values using the most recent information less than the new data (see details).
• cond_mean: an optional matrix of the filtered conditional mean values.
• ...: additional arguments for future expansion.

Returns

A cgarch.estimate or dcc.estimate object with updated information. All values in the object are updated with the exception of the hessian and scores which remain at their estimation set values.

Details

The method filters new data and updates the object with this new information so that it can be called recursively as new data arrives. The update argument allows values such as the intercept matrices and transformation estimates (for the spd and empirical methods) in the dynamic case, and the constant correlation in the constant case, to use information up to and include time T, where T is the time stamp just preceding the new y timestamps. In this way, the filter method can be called recursively and the user can selectively choose to either use the updating scheme or use the original estimated values. Whatever the case, this ensures that there is no look-ahead bias when filtering new information.