R help - Apr 2000 - (1) arima.mle implementation; (2) r-newbie forum

Has arima.mle (MASS, Ch.15, p.464) been implemented in R?

[A search through contributed packages and R-FAQ suggests not,
and I don't think function 'lme' from package 'nlme' would
accomplish the same sort of thing, although it permits
correlated errors.  A search of the CRAN site shows this 
question has been asked some time ago, and it was suggested
that Paul Gilbert's DSE library might offer comparable 
functions.  Is this the way to go?]

Using R 1.0.0 for Win9x. Moving to Linux in September :)

Incidentally, R is great, and so is this forum.  As the user
base is expanding, it might be a good idea to start a second
help forum where newbies can discuss "stupid questions" amongst
ourselves.  I think there are alot more R-related questions out
there than are currently being asked.  And even now many of
the questions being posted have as much to do with the use of
statistics as with the use of R proper.  Having another
forum (r-newuser? r-statshelp?) might free up core team members
from having to answer those trivial questions that experienced
users can easily handle and questions that are only tangentially
related to R.

[For example I would like to know, but am hesitant to ask:
1. How does one interpret the confidence limits computed by
spec.pgram and automatically plotted by plot.spec?
2. What is the most elegant way to generate a vector of
correlated random numbers of arbitrary distribution, mean, 
variance and correlation structure? (eg. pink noise).]

Sincerely,

Barry J. Cooke                      Current mailing address:
Ph.D. Candidate                     #219 - 1810 NW 23 Blvd 
Environmental Biology and Ecology   Gainesville, Florida, USA 
Department of Biological Sciences   32605 
University of Alberta
Edmonton, AB, T6G 2E9               http://www.ualberta.ca/~bcooke 
bcooke at gpu.srv.ualberta.ca

-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
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
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

On Sun, 9 Apr 2000, Barry Cooke wrote:
> Has arima.mle (MASS, Ch.15, p.464) been implemented in R?
> 
> [A search through contributed packages and R-FAQ suggests not,
> and I don't think function 'lme' from package 'nlme'
would
> accomplish the same sort of thing, although it permits
> correlated errors.  A search of the CRAN site shows this 
> question has been asked some time ago, and it was suggested
> that Paul Gilbert's DSE library might offer comparable 
> functions.  Is this the way to go?]
Look at the on-line V&R `R Complements' for how to do things in MASS in
R.
See arima0 in library ts (which is a standard package).
> Using R 1.0.0 for Win9x. Moving to Linux in September :)
> 
> Incidentally, R is great, and so is this forum.  As the user
> base is expanding, it might be a good idea to start a second
> help forum where newbies can discuss "stupid questions" amongst
> ourselves.  I think there are alot more R-related questions out
> there than are currently being asked.  And even now many of
> the questions being posted have as much to do with the use of
> statistics as with the use of R proper.  Having another
> forum (r-newuser? r-statshelp?) might free up core team members
> from having to answer those trivial questions that experienced
> users can easily handle and questions that are only tangentially
> related to R.
We would be very happy for `experienced users' to answer them on r-help.
> [For example I would like to know, but am hesitant to ask:
> 1. How does one interpret the confidence limits computed by
> spec.pgram and automatically plotted by plot.spec?
(pointwise? Sorry, I miss the point here. 
You found package ts, so how did you miss arima0?)
> 2. What is the most elegant way to generate a vector of
> correlated random numbers of arbitrary distribution, mean, 
> variance and correlation structure? (eg. pink noise).]
Well, you can't have arbitrary distribution and covariance structure.
There are theoretical constraints on the correlation between two rv's,
even. For normal distributions, see section 4.5 of BDR (1987) Stochastic
Simulation and choose the method that takes your fancy.

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272860 (secr)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595

-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
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
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

>Has arima.mle (MASS, Ch.15, p.464) been implemented in R?
>...
>it was suggested
>that Paul Gilbert's DSE library might offer comparable
>functions.  Is this the way to go?]
(I am hoping to soon have a new version of DSE in the development area
of CRAN.)

My library has several multivariate time series estimation techniques,
but maximum likelihood is not currently working and probably will not be
for some time to come. For a long time I thought maximum likelihood was
the best way to estimate. But, while discussing rather unimpressive
results from simulation tests, a friend remarked that "every knows mle
gives good in sample fit at the expense of out of sample prediction."
Whether everyone knows that I rather doubt, but in any case it is
certainly consistence with my experience and is the main reason I do not
expect to fix maximum likelihood estimation in DSE anytime soon. (Not to
mention that ml estimation of multivariate time series models takes days
on fast computers and one is never really sure about convergence.)

Paul Gilbert

-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
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
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._