this is kind of a rejoinder to our posting to r-devel "request for comments --- package "distr" --- S4 Classes for Distributions" dating from 01-31-04/ 05:11:50 EST./
To start, we want to thank the contributors of comments and suggestions, which have been quite helpful to us.
Also, we want to apologize for being somewhat lengthy in our replies; we hope you could cope with this.
Some specific questions emerged from this discussion:
(1) Naming of "private" slots: We agree with a proposal by Duncan Murdoch
[off-list] to use longer names in order to inhibit misuse of the @-operator
(2) Naming of the accesors r,d,p,q: We tentatively would keep our naming as it makes apparent their origin.
(3) Question of functions as slots: We hope to have convinced you that our solution has some advantages in connection with transformations of distributions.
(4) Redundant saving of values of slots (in case of the parameter of a distribution --- a topic mentioned by Duncan Murdoch [off-list]):
YES we do have some redundant saving of values. BUT if inspection / modification of slots is done only by accessor/ replacement functions we take care of consistency. We admit a certain inefficiency in that we replace the whole object (including the validity check) when modifying parameter slots; --- it just seemed much easier to realize and on the other hand, as no too large structures are part of a distribution object, we see this inefficiency only as a minor disadvantage.
(5) Multivariate distributions They could easily be included to our package, but haven't been yet --- we simply do not consider ourselves experts enough to do this job.
It would be nice if you could let us know your opinion / some suggestions to these issues.
Thank you once again for your attention.
Peter Ruckdeschel, Matthias Kohl, Thomas Stabla, Florian Camphausen. ------------------------------------------------------------------------------------------ The http-reference for package "distr"
http://www.uni-bayreuth.de/departments/math/org/mathe7/DISTR/ -------------------------------------------------------------------------------------------
______________________________________________ [EMAIL PROTECTED] mailing list https://www.stat.math.ethz.ch/mailman/listinfo/r-devel
