tscokurt.tsmarch function

Cokurtosis Extractor

Extracts the conditional cokurtosis matrices.

## S3 method for class 'gogarch.estimate'
tscokurt(
  object,
  index = NULL,
  distribution = FALSE,
  standardize = TRUE,
  folded = TRUE,
  ...
)

## S3 method for class 'gogarch.predict'
tscokurt(
  object,
  index = NULL,
  distribution = FALSE,
  standardize = TRUE,
  folded = TRUE,
  ...
)

## S3 method for class 'gogarch.simulate'
tscokurt(
  object,
  index = NULL,
  distribution = FALSE,
  standardize = TRUE,
  folded = TRUE,
  ...
)

Arguments

• object: an object class from one of the models in the package.
• index: the time index (integer) from which to extract a subset of the cokurtosis array rather than the whole time series.
• distribution: whether to return the full simulated cokurtosis distribution for the predicted and simulated objects, else the average cokurtosis across each horizon.
• standardize: whether to standardize the 4th co-moment so that it represents the cokurtosis.
• folded: whether to return the result as a folded or unfolded array. The folded array is n_series x n_series x n_series x n_series x horizon (x simulation if predicted or simulated object). The unfolded array is a n_series x (n_series^3) x horizon array. Calculations such as weighted co-moments are based on the unfolded array using the Kronecker operator.
• ...: none

Returns

the cokurtosis (see details).

Details

The calculation of the cokurtosis array from the independent factors is very expensive in terms of memory footprint as well as computation time. While it does take advantage of multiple threads if required (see setThreadOptions), in the case of many series this will quickly become difficult for systems low RAM. Because of this, there is the option to extract a specific point in time output using the index

argument.

Author(s)

Alexios Galanos