DSE package for multi-variate time series

[Paul Gilbert]

A new version of my DSE package for multi-variate time series analysis
is
now available at <www.bank-banque-canada.ca/pgilbert>. I am trying to
sort
out some minor glitches with the new R 0.90 graphics before I submit it
to CRAN. Much of the underlying code has been re-worked and reorganized
in the
new version. In particular, the use of model and data constructors has
been
more formalized and the internal structure of these objects, while still

documented, is considered "opaque."

I have separated out a "syskern" package which tries to provide a kernel
of
routines to isolate OS differences and a few Splus and R differences.
For most
people, the most interesting part of this is probably the approach to
RNG,
which provides a convenient way to generate the same random experiments
in R
and Splus.

I have also separated out a "tframe" package which provides a kernel of
routines for programming time series methods. These allow a programmer
to
write most code in a way that is independent of the time representation
(i.e.
the class of the time series data). So, for example, the use of tsp() is
avoided
because it is specific to certain classes of time series.

The changes to the underlying structure necessitated a substantial
re-write of
the user's guide, so I am taking the opportunity to bring the guide
up-to-date
with respect to R. A draft of the guide is available at the above web
site in
postscript and pdf files. The first nine sections, which cover material
in the previous version of the guide, are now mostly complete and I hope

correct. I would certainly appreciate comments. Some later sections,
which
cover new material, are still missing or incomplete.

The help has largely been converted to integrate with R's help. I have
not yet worked through all the examples, many of which were written with
fictitious data in mind, so it will be some time before "R CMD check
dse" works, however, most of my tests work in both Linux and Solaris.
These include:

   random.number.test()
   tframe.function.tests()
   dse1.function.tests()
   dse2.function.tests()
   dse3.function.tests()
   dse4.function.tests()
   guide.example.tests.part1()
   guide.example.tests.part2()

which cover a large part of the package. The comparison tolerances on
some of
the tests had to be relaxed in order to pass with R on Linux and R and
Splus 3.3
in Solaris. I would appreciate feedback about how the tests work on
other
platforms.

Paul Gilbert



-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-announce 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-announce-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._