				-*- org -*-
* TODO WhittleEst() not converging [-> f.am31 in man/WhittleEst.Rd ]: *do keep* the warning
 in the resulting object and *do* show that with summary() / print()
* TODO Compare with   library( fracdiff )  -- and 'ts2test' :
    --> partly happened in "Semesterarbeit" by Thomas Grischott
	(-> /u/maechler/save/papers/RGrischott-LongMem.pdf
	  and /sfs/a/A/SfS-students/groman/RAusw/ )
  --> Completely merge with the also unreleased package ../ts2test !!

* TODO fixup ./man/specFGN.Rd  {*math* (\eqn{LaTeX}..) formula, MM's story,..}
  ---> more clearly explain the approximation and 'k.approx', etc.
  ---> relying on Paxson's paper (for now in ./Paxson-1997_FGN-CCR.pdf

* TODO Martin Maechler thinks that Paxson's approach should/could be replaced by
  a *better* one, using the Euler-Maclaurin formula for the sum.
  ( --> inst/doc/Euler-MacLaurin.tex in MM's path )

* TODO provide  logLik() methods .. For WhittleEst(), should be straightforward,
  using Jan's book and Qeta().

* TODO Think about adding Graf's  *robust* estimators ("HUBINC" et al.).
  ---> package ../ts2test/ which has M() !

* TODO specFGN(eta, m, nsum)  becomes quite slow for "large" m
  (e.g m = 2^16; the for loop takes time).
  Should do this in C code!
  ==> even for n=1000, this is too slow!

  >>>>>> Diethelm Wuertz got code from Taqqu, to do this *much* faster,
  ---> see 'fArma' Rmetric package
  (MM: ~/R/D/R-forge/Rmetrics/pkg/fArma/R/whittle.R
