This method minimizes an approximate log-likelihood function (applied to the spectral density) to obtain an estimate of the parameters of a process. It is parametric and does not produce a graphical output.

A more detailed description.


There are several options for using the Whittle estimator. Some are described below.

  1. The data is here called xinput.
  2. One can optionally subdivide the series into subseries (nsub, default is 1).
  3. One can output the periodogram (spec, default = NO).
  4. One can ouput intermediate minimization results (out, default = YES).
  5. The model can be either "farima" or "fgn".
  6. If it is "farima", pp and qq set the order of the autoregressive and moving average parts respectively.
  7. h is the starting value of H for the minimization procedure.
  8. ar and ma are the vectors corresponding to the starting values of the other parameters. (Length of vectors should be the same as pp and qq.
  9. To use the Aggregated Whittle method, first aggregate the data, and then use the FGN model in the Whittle estimator.