Patch to add Beta binomial distribution. Mentor needed!

[Joan Maspons]

Hello,

I implemented the Beta binomial distribution following the patterns of the
binomial distribution code and inspired by JAGS' code [1]. I have studied
the code carefully but it's my first run in the R internals.

Can somebody review the code and if everything it's ok commit to the
repository?

[1]
http://mcmc-jags.hg.sourceforge.net/hgweb/mcmc-jags/mcmc-jags/file/15af65a4be29/src/modules/bugs/distributions/DBetaBin.cc


--
Joan Maspons
CREAF (Centre de Recerca Ecològica i Aplicacions Forestals)
Universitat Autònoma de Barcelona, 08193 Bellaterra (Barcelona), Catalonia
Tel [phone]            [hidden email]
http://www.creaf.uab.cat

______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel

[Christophe Dutang1]

Dear Joan,

Are you aware of this page http://cran.r-project.org/web/views/Distributions.html ?

If you want to contribute to R, you should write a package and submit it to CRAN. See http://cran.r-project.org/doc/manuals/R-exts.html

Regards

Christophe

--
Christophe Dutang
Ph.D. student at ISFA, Lyon, France
website: http://dutangc.free.fr

Le 22 mai 2012 à 12:14, Joan Maspons a écrit :

> Hello,
>
> I implemented the Beta binomial distribution following the patterns of the
> binomial distribution code and inspired by JAGS' code [1]. I have studied
> the code carefully but it's my first run in the R internals.
>
> Can somebody review the code and if everything it's ok commit to the
> repository?
>
> [1]
> http://mcmc-jags.hg.sourceforge.net/hgweb/mcmc-jags/mcmc-jags/file/15af65a4be29/src/modules/bugs/distributions/DBetaBin.cc
>
>
> --
> Joan Maspons
> CREAF (Centre de Recerca Ecològica i Aplicacions Forestals)
> Universitat Autònoma de Barcelona, 08193 Bellaterra (Barcelona), Catalonia
> Tel [phone]            [hidden email]
> http://www.creaf.uab.cat
> ______________________________________________
> [hidden email] mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel

______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel

[Joan Maspons]

Hello,

2012/5/22 Christophe Dutang <[hidden email]>

> Dear Joan,
>
> Are you aware of this page
> http://cran.r-project.org/web/views/Distributions.html ?
>

Yes, I remember
https://stat.ethz.ch/pipermail/r-devel/2012-March/063552.html


> If you want to contribute to R, you should write a package and submit it
> to CRAN. See http://cran.r-project.org/doc/manuals/R-exts.html
>

It's necessary to develop a package before to add the functions to R-base?
I just want to add beta binomial and beta negative binomial distributions
which from my point of view are not more strange than distributions
included in R-base. I think they are quite common in order to simulate a
binomial or negative binomial process in an stochastic environment.

In the other hand I have no experience in writing packages so it would take
me more time.

Yours,
Joan

Regards

>
> Christophe
>
> --
> Christophe Dutang
> Ph.D. student at ISFA, Lyon, France
> website: http://dutangc.free.fr
>
> Le 22 mai 2012 à 12:14, Joan Maspons a écrit :
>
> > Hello,
> >
> > I implemented the Beta binomial distribution following the patterns of
> the
> > binomial distribution code and inspired by JAGS' code [1]. I have studied
> > the code carefully but it's my first run in the R internals.
> >
> > Can somebody review the code and if everything it's ok commit to the
> > repository?
> >
> > [1]
> >
> http://mcmc-jags.hg.sourceforge.net/hgweb/mcmc-jags/mcmc-jags/file/15af65a4be29/src/modules/bugs/distributions/DBetaBin.cc
> >
> >
> > --
> > Joan Maspons
> > CREAF (Centre de Recerca Ecològica i Aplicacions Forestals)
> > Universitat Autònoma de Barcelona, 08193 Bellaterra (Barcelona),
> Catalonia
> > Tel [phone]            [hidden email]
> > http://www.creaf.uab.cat
> > ______________________________________________
> > [hidden email] mailing list
> > https://stat.ethz.ch/mailman/listinfo/r-devel
>
>

--
Joan Maspons
CREAF (Centre de Recerca Ecològica i Aplicacions Forestals)
Universitat Autònoma de Barcelona, 08193 Bellaterra (Barcelona), Catalonia
Tel [phone]            [hidden email]
http://www.creaf.uab.cat

        [[alternative HTML version deleted]]


______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel

[Joshua Ulrich]

On Tue, May 22, 2012 at 8:27 AM, Joan Maspons <[hidden email]> wrote:
> Hello,
>
> 2012/5/22 Christophe Dutang <[hidden email]>
>
>> Dear Joan,
>>
<snip>

>> If you want to contribute to R, you should write a package and submit it
>> to CRAN. See http://cran.r-project.org/doc/manuals/R-exts.html
>>
>
> It's necessary to develop a package before to add the functions to R-base?
> I just want to add beta binomial and beta negative binomial distributions
> which from my point of view are not more strange than distributions
> included in R-base. I think they are quite common in order to simulate a
> binomial or negative binomial process in an stochastic environment.
>

In order to add functions to the main R distribution, you have to
convince someone on R-core to maintain the code.

> In the other hand I have no experience in writing packages so it would take
> me more time.
>
The actuar package contains many additional distributions.  You might
see if the maintainer(s) would be interested in incorporating your
code (if they don't have that functionality already).

> Yours,
> Joan
>
> Regards
>>
>> Christophe
>>
>> --
>> Christophe Dutang
>> Ph.D. student at ISFA, Lyon, France
>> website: http://dutangc.free.fr
>>
>> Le 22 mai 2012 à 12:14, Joan Maspons a écrit :
>>
>> > Hello,
>> >
>> > I implemented the Beta binomial distribution following the patterns of
>> the
>> > binomial distribution code and inspired by JAGS' code [1]. I have studied
>> > the code carefully but it's my first run in the R internals.
>> >
>> > Can somebody review the code and if everything it's ok commit to the
>> > repository?
>> >
>> > [1]
>> >
>> http://mcmc-jags.hg.sourceforge.net/hgweb/mcmc-jags/mcmc-jags/file/15af65a4be29/src/modules/bugs/distributions/DBetaBin.cc
>> >
>> >
>> > --
>> > Joan Maspons
>> > CREAF (Centre de Recerca Ecològica i Aplicacions Forestals)
>> > Universitat Autònoma de Barcelona, 08193 Bellaterra (Barcelona),
>> Catalonia
>> > Tel [phone]            [hidden email]
>> > http://www.creaf.uab.cat
>> > ______________________________________________
>> > [hidden email] mailing list
>> > https://stat.ethz.ch/mailman/listinfo/r-devel
>>
>>
>
>
> --
> Joan Maspons
> CREAF (Centre de Recerca Ecològica i Aplicacions Forestals)
> Universitat Autònoma de Barcelona, 08193 Bellaterra (Barcelona), Catalonia
> Tel [phone]            [hidden email]
> http://www.creaf.uab.cat
>
>        [[alternative HTML version deleted]]
>
>
> ______________________________________________
> [hidden email] mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel
>

______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel

[Tim Triche, Jr.]

it's already in the VGAM package

http://www.stat.auckland.ac.nz/~yee/VGAM/



On Tue, May 22, 2012 at 7:59 AM, Joshua Ulrich <[hidden email]>wrote:

> On Tue, May 22, 2012 at 8:27 AM, Joan Maspons <[hidden email]>
> wrote:
> > Hello,
> >
> > 2012/5/22 Christophe Dutang <[hidden email]>
> >
> >> Dear Joan,
> >>
> <snip>
> >> If you want to contribute to R, you should write a package and submit it
> >> to CRAN. See http://cran.r-project.org/doc/manuals/R-exts.html
> >>
> >
> > It's necessary to develop a package before to add the functions to
> R-base?
> > I just want to add beta binomial and beta negative binomial distributions
> > which from my point of view are not more strange than distributions
> > included in R-base. I think they are quite common in order to simulate a
> > binomial or negative binomial process in an stochastic environment.
> >
> In order to add functions to the main R distribution, you have to
> convince someone on R-core to maintain the code.
>
> > In the other hand I have no experience in writing packages so it would
> take
> > me more time.
> >
> The actuar package contains many additional distributions.  You might
> see if the maintainer(s) would be interested in incorporating your
> code (if they don't have that functionality already).
>
> > Yours,
> > Joan
> >
> > Regards
> >>
> >> Christophe
> >>
> >> --
> >> Christophe Dutang
> >> Ph.D. student at ISFA, Lyon, France
> >> website: http://dutangc.free.fr
> >>
> >> Le 22 mai 2012 à 12:14, Joan Maspons a écrit :
> >>
> >> > Hello,
> >> >
> >> > I implemented the Beta binomial distribution following the patterns of
> >> the
> >> > binomial distribution code and inspired by JAGS' code [1]. I have
> studied
> >> > the code carefully but it's my first run in the R internals.
> >> >
> >> > Can somebody review the code and if everything it's ok commit to the
> >> > repository?
> >> >
> >> > [1]
> >> >
> >>
> http://mcmc-jags.hg.sourceforge.net/hgweb/mcmc-jags/mcmc-jags/file/15af65a4be29/src/modules/bugs/distributions/DBetaBin.cc
> >> >
> >> >
> >> > --
> >> > Joan Maspons
> >> > CREAF (Centre de Recerca Ecològica i Aplicacions Forestals)
> >> > Universitat Autònoma de Barcelona, 08193 Bellaterra (Barcelona),
> >> Catalonia
> >> > Tel [phone]            [hidden email]
> >> > http://www.creaf.uab.cat
> >> > ______________________________________________
> >> > [hidden email] mailing list
> >> > https://stat.ethz.ch/mailman/listinfo/r-devel
> >>
> >>
> >
> >
> > --
> > Joan Maspons
> > CREAF (Centre de Recerca Ecològica i Aplicacions Forestals)
> > Universitat Autònoma de Barcelona, 08193 Bellaterra (Barcelona),
> Catalonia
> > Tel [phone]            [hidden email]
> > http://www.creaf.uab.cat
> >
> >        [[alternative HTML version deleted]]
> >
> >
> > ______________________________________________
> > [hidden email] mailing list
> > https://stat.ethz.ch/mailman/listinfo/r-devel
> >
>
> ______________________________________________
> [hidden email] mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel
>

--
*A model is a lie that helps you see the truth.*
*
*
Howard Skipper<http://cancerres.aacrjournals.org/content/31/9/1173.full.pdf>

        [[alternative HTML version deleted]]


______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel

[bbolker]

Tim Triche, Jr. <tim.triche <at> gmail.com> writes:

 
> it's already in the VGAM package
>
> http://www.stat.auckland.ac.nz/~yee/VGAM/
>

[various snippage]

also see

library("sos")
findFn("dbetabin*")

and

http://stackoverflow.com/questions/8065835/
   proposing-feature-requests-to-the-r-core-team/8066062#8066062
[url broken]

  for some context on why you should build a package instead
of trying to get a patch accepted in base R ...

  Ben Bolker

 

> > On Tue, May 22, 2012 at 8:27 AM, Joan Maspons
> > wrote:
> > > Hello,
> > >
> > > 2012/5/22 Christophe Dutang <dutangc <at> gmail.com>
> > >
> > >> Dear Joan,
> > >>
> > <snip>
> > >> If you want to contribute to R, you should write a package and submit it
> > >> to CRAN. See http://cran.r-project.org/doc/manuals/R-exts.html
> > >>
> > >
> > > It's necessary to develop a package before to add the functions to
> > R-base?
> > > I just want to add beta binomial and beta negative binomial distributions
> > > which from my point of view are not more strange than distributions
> > > included in R-base. I think they are quite common in order to simulate a
> > > binomial or negative binomial process in an stochastic environment.
> > >
> > In order to add functions to the main R distribution, you have to
> > convince someone on R-core to maintain the code.

______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel

[Joan Maspons]

Hi,
I filled a bug [1] and I tried to build a package without success. If there
are no news in a while I'll try some to convince some package developers to
add this distribution

[1] https://bugs.r-project.org/bugzilla3/show_bug.cgi?id=14936

Thanks for the feedback,
Joan

2012/5/22 Ben Bolker <[hidden email]>

>
> [various snippage]
>
> also see
>
> library("sos")
> findFn("dbetabin*")
>
> and
>
> http://stackoverflow.com/questions/8065835/
>   proposing-feature-requests-to-the-r-core-team/8066062#8066062
> [url broken]
>
>  for some context on why you should build a package instead
> of trying to get a patch accepted in base R ...
>
>  Ben Bolker
>
>
> > > On Tue, May 22, 2012 at 8:27 AM, Joan Maspons
> > > wrote:
> > > > Hello,
> > > >
> > > > 2012/5/22 Christophe Dutang <dutangc <at> gmail.com>
> > > >
> > > >> Dear Joan,
> > > >>
> > > <snip>
> > > >> If you want to contribute to R, you should write a package and
> submit it
> > > >> to CRAN. See http://cran.r-project.org/doc/manuals/R-exts.html
> > > >>
> > > >
> > > > It's necessary to develop a package before to add the functions to
> > > R-base?
> > > > I just want to add beta binomial and beta negative binomial
> distributions
> > > > which from my point of view are not more strange than distributions
> > > > included in R-base. I think they are quite common in order to
> simulate a
> > > > binomial or negative binomial process in an stochastic environment.
> > > >
> > > In order to add functions to the main R distribution, you have to
> > > convince someone on R-core to maintain the code.
>
> ______________________________________________
> [hidden email] mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel
>

--
Joan Maspons
CREAF (Centre de Recerca Ecològica i Aplicacions Forestals)
Universitat Autònoma de Barcelona, 08193 Bellaterra (Barcelona), Catalonia
Tel [phone]            [hidden email]
http://www.creaf.uab.cat

        [[alternative HTML version deleted]]


______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel

[Tim Triche, Jr.]

It's in VGAM, remember?  Not a bug.



On Fri, Jun 1, 2012 at 10:38 AM, Joan Maspons <[hidden email]>wrote:

> Hi,
> I filled a bug [1] and I tried to build a package without success. If there
> are no news in a while I'll try some to convince some package developers to
> add this distribution
>
> [1] https://bugs.r-project.org/bugzilla3/show_bug.cgi?id=14936
>
> Thanks for the feedback,
> Joan
>
> 2012/5/22 Ben Bolker <[hidden email]>
>
> >
> > [various snippage]
> >
> > also see
> >
> > library("sos")
> > findFn("dbetabin*")
> >
> > and
> >
> > http://stackoverflow.com/questions/8065835/
> >   proposing-feature-requests-to-the-r-core-team/8066062#8066062
> > [url broken]
> >
> >  for some context on why you should build a package instead
> > of trying to get a patch accepted in base R ...
> >
> >  Ben Bolker
> >
> >
> > > > On Tue, May 22, 2012 at 8:27 AM, Joan Maspons
> > > > wrote:
> > > > > Hello,
> > > > >
> > > > > 2012/5/22 Christophe Dutang <dutangc <at> gmail.com>
> > > > >
> > > > >> Dear Joan,
> > > > >>
> > > > <snip>
> > > > >> If you want to contribute to R, you should write a package and
> > submit it
> > > > >> to CRAN. See http://cran.r-project.org/doc/manuals/R-exts.html
> > > > >>
> > > > >
> > > > > It's necessary to develop a package before to add the functions to
> > > > R-base?
> > > > > I just want to add beta binomial and beta negative binomial
> > distributions
> > > > > which from my point of view are not more strange than distributions
> > > > > included in R-base. I think they are quite common in order to
> > simulate a
> > > > > binomial or negative binomial process in an stochastic environment.
> > > > >
> > > > In order to add functions to the main R distribution, you have to
> > > > convince someone on R-core to maintain the code.
> >
> > ______________________________________________
> > [hidden email] mailing list
> > https://stat.ethz.ch/mailman/listinfo/r-devel
> >
>
>
>
> --
> Joan Maspons
> CREAF (Centre de Recerca Ecològica i Aplicacions Forestals)
> Universitat Autònoma de Barcelona, 08193 Bellaterra (Barcelona), Catalonia
> Tel [phone]            [hidden email]
> http://www.creaf.uab.cat
>
>         [[alternative HTML version deleted]]
>
>
> ______________________________________________
> [hidden email] mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel
>
>

--
*A model is a lie that helps you see the truth.*
*
*
Howard Skipper<http://cancerres.aacrjournals.org/content/31/9/1173.full.pdf>

        [[alternative HTML version deleted]]


______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel

[Joan Maspons]

2012/6/1 Tim Triche, Jr. <[hidden email]>

> It's in VGAM, remember?  Not a bug.
>
> --
> *A model is a lie that helps you see the truth.*
> *
> *
> Howard Skipper<http://cancerres.aacrjournals.org/content/31/9/1173.full.pdf>
>
>
Yes, it's a wish, not a bug. I would like a faster implementation in c and
including quantile function. pbetabin calls qbetabin and sums from 0 to q
using a loop. It can be avoided using genhypergeo function from hypergeo
package [1] but it adds a dependency and for some parameters it doesn't
converge.

In the other hand I think that these functions should be implemented in a
more neutral place to avoid that packages interested in this distribution
have to depend on VGAM. I'm also interested in the Beta negative binomial
distribution which I think it's not implemented (this will be the second
round). And last but not least I want to learn.

Yours,
--
Joan Maspons
CREAF (Centre de Recerca Ecològica i Aplicacions Forestals)
Universitat Autònoma de Barcelona, 08193 Bellaterra (Barcelona), Catalonia
Tel [phone]            [hidden email]
http://www.creaf.uab.cat


[1] https://stat.ethz.ch/pipermail/r-devel/2012-March/063560.html

        [[alternative HTML version deleted]]


______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel

[bbolker]

Joan Maspons <j.maspons <at> creaf.uab.cat> writes:

>
> 2012/6/1 Tim Triche, Jr. <tim.triche <at> gmail.com>
>
> > It's in VGAM, remember?  Not a bug.
> >
> > --
> > *A model is a lie that helps you see the truth.*
> > *
> > *
> > Howard Skipper<http://cancerres.aacrjournals.org/content/31/9/1173.full.pdf>
> >
> >
> Yes, it's a wish, not a bug. I would like a faster implementation in c and
> including quantile function. pbetabin calls qbetabin and sums from 0 to q
> using a loop. It can be avoided using genhypergeo function from hypergeo
> package [1] but it adds a dependency and for some parameters it doesn't
> converge.
>
> In the other hand I think that these functions should be implemented in a
> more neutral place to avoid that packages interested in this distribution
> have to depend on VGAM. I'm also interested in the Beta negative binomial
> distribution which I think it's not implemented (this will be the second
> round). And last but not least I want to learn.

  I really think your chances of getting this wish accepted are tiny.
If you want to avoid additional dependencies, you have at least the
*legal* right to take any piece of code you want from any open-source
(BSD, GPL, ...) package and package it your own little package
(provided it follows the rules of the original license) that
includes only the stuff you need, without any other dependencies.  It's
polite to ask for permission and acknowledge the original authors ...

  Ben Bolker

______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel

[Prof Brian Ripley]

On 05/06/2012 09:25, Ben Bolker wrote:

> Joan Maspons<j.maspons<at>  creaf.uab.cat>  writes:
>
>>
>> 2012/6/1 Tim Triche, Jr.<tim.triche<at>  gmail.com>
>>
>>> It's in VGAM, remember?  Not a bug.
>>>
>>> --
>>> *A model is a lie that helps you see the truth.*
>>> *
>>> *
>>> Howard Skipper<http://cancerres.aacrjournals.org/content/31/9/1173.full.pdf>
>>>
>>>
>> Yes, it's a wish, not a bug. I would like a faster implementation in c and
>> including quantile function. pbetabin calls qbetabin and sums from 0 to q
>> using a loop. It can be avoided using genhypergeo function from hypergeo
>> package [1] but it adds a dependency and for some parameters it doesn't
>> converge.
>>
>> In the other hand I think that these functions should be implemented in a
>> more neutral place to avoid that packages interested in this distribution
>> have to depend on VGAM. I'm also interested in the Beta negative binomial
>> distribution which I think it's not implemented (this will be the second
>> round). And last but not least I want to learn.
>
>    I really think your chances of getting this wish accepted are tiny.
> If you want to avoid additional dependencies, you have at least the
> *legal* right to take any piece of code you want from any open-source
> (BSD, GPL, ...) package and package it your own little package
> (provided it follows the rules of the original license) that
> includes only the stuff you need, without any other dependencies.  It's
> polite to ask for permission and acknowledge the original authors ...

It is not just polite.  You need to comply with copyright law, and it is
very likely that requires acknowledgement.  So do CRAN's policies (and
we've removed several packages already this year for misrepresentation
of the work of others).

>    Ben Bolker
>
> ______________________________________________
> [hidden email] mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel


--
Brian D. Ripley,                  [hidden email]
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  [phone] (self)
1 South Parks Road,                     [phone] (PA)
Oxford OX1 3TG, UK                Fax:  [phone]

______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel

[Joan Maspons]

Hello,

I'm trying to implement the Beta binomial distribution as a R package.
I studied R-ext manual but still there are some problems. The code in
C seems to work except for the rbetabinom function which is not very
random (it always returns the same values for the same parameters). I
have tried to use GetRNGstate(); and PutRNGstate(); but it give a SEG
FAULT when I call GetRNGstate. You can try the code like this:

cd /path/to/src
cmake ./CMakeLists.txt
make
./r-package


I can build and install the package but when I call the functions from
R they always return NULL except for rbetabinom which remains hanged
indefinitely.

Any advise?

--
Joan Maspons
CREAF (Centre de Recerca Ecològica i Aplicacions Forestals)
Universitat Autònoma de Barcelona, 08193 Bellaterra (Barcelona), Catalonia
Tel [phone]            [hidden email]
http://www.creaf.uab.cat

______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel