Dirk Eddelbuettel edd at debian.org
Sun Dec 31 20:12:04 CET 2000

  "Jason" == Jason Turner <jasont at indigoindustrial.co.nz> writes:
  Jason> You can also use the fracdiff package for a fractionally differenced
  Jason> arima model (max. likelihood estimation of parameters).

Fracdiff's covariance matrix always leave me puzzled. Consider this (slightly
modified) 'example(fracdiff)' session:

> library(fracdiff)
> ts.test <- fracdiff.sim( 5000, ar = .2, ma = -.4, d = .3)
> .Random.seed
[1]          1 1866294289  525838217
>  fd.obj <- fracdiff( ts.test$series, nar = length(ts.test$ar), nma = length(ts.test$ma))
Warning message:
unable to compute correlation matrix in: switch(temp$info, warning("warning in gamma function"), warning("singular Hessian"),
>  fd.obj$hessian.dpq
                d       ar1      ma1
d   -2.926964e+11 -4251.568 4249.104
ar1 -4.251568e+03     0.000    0.000
ma1  4.249104e+03     0.000    0.000      

Quoting from Venables and Ripley (S Programming, p 168): "no estimate is
complete without some estimate of its uncertainty". 

That would leave fracdiff incomplete, or even broken, wouldn't it?

Cheers, Dirk

According to the latest figures, 43% of all statistics are totally worthless.
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch

More information about the R-help mailing list