[Rd] [R] Problems with ARIMA models?

Wacek Kusnierczyk Waclaw.Marcin.Kusnierczyk at idi.ntnu.no
Thu Feb 26 09:49:46 CET 2009

that's interesting, marie.

on his webpage [1], david s. stoffer provides concrete arguments for why
the r arima function in the stats package should be considered to
produce misleading, if not just wrong results. 

david (cc:ed for reference) reported the issue in october 2005, bug
report #8231 [2], and then again in april 2006, same bug report id [3]. 
a search through the maling list's archives reveals that there are only
two messages matching the pattern '8231', namely the two mentioned
above.  that is, there has been *no public response* posted.

after a few lessons, i have learned that you can't demand anything from
r developers (which isn't quite accurate:  you can demand, but they're
likely to ignore you), so it's hard to demand that this issue be
appropriately addressed.  however, one can, and should, expect that such
bug reports, correct or not, are addressed in public, leaving no doubt
as to the accuracy of the design and implementation of any functionality
in r.

responding to bug reports is an important part of the open source
culture (in this particular case, it's not really about sources) which r
folks do not hesitate to refer to.  it is surprising that david's
concrete and polite (please prove me wrong) post hasn't received the
treatment it deserved.  it hasn't any! 

a brief look at some of the respective sources:

    svn log http://svn.r-project.org/R/trunk/src/library/stats/R/arima.R
    svn log http://svn.r-project.org/R/trunk/src/library/stats/src/arima.c

hints that they've been edited mostly by one single developer.  if the
responsibility, as it seems, is not diffused among many, why wouldn't
the author provide a clear and convincing answer to david's comments? 
as far as i can see, david has sufficient expertise in both the
particular field and r in general, not to be treated as a naive user
asking dumb questions (see fortune('dumb'), for example).

is ignoring concrete critique part of the official strategy?  can't see
this documented on r's website.


[1] http://www.stat.pitt.edu/stoffer/tsa2/Rissues.htm
[2] http://tolstoy.newcastle.edu.au/R/devel/05/10/2885.html
[3] http://tolstoy.newcastle.edu.au/R/devel/06/04/4756.html

Marie Sivertsen wrote:
> Dear R,
> I have find a website where they report problem with ARIMA models in R.  I
> run the examples there and they give result as shown on the website.  Does
> this mean that nothing has corrected in R?  Maybe you not have seen the
> page, but the author said he contacted you.
> Here is the URL: http://www.stat.pitt.edu/stoffer/tsa2/Rissues.htm
> I like to know your opinion.
> Mvh.
> Marie
> 	[[alternative HTML version deleted]]
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.

More information about the R-devel mailing list