[GenABEL-dev] Implementing p-value output in ProbABEL

Yurii Aulchenko yurii.aulchenko at gmail.com
Sun May 26 12:25:05 CEST 2013


On Sun, May 26, 2013 at 4:34 PM, L.C. Karssen <lennart at karssen.org> wrote:

> Hi Yurii,
>
> Thanks for your insights!
>
> On 26-05-13 09:22, Yurii Aulchenko wrote:
> > R: that may be too much to ask to install R as a 'library' :) however,
> > similarly to boost, this may provide with more opportunities in the
> future
> >
> > GSL: to my experience should work across major platforms (at least I
> > remember compiling MixABEL for Win, and I am using MixABEL for Mac);
> > using GSL in MixABEL worked rather straightforwardly
>
> Good to hear that. I do find GSL's matrix/vector code a hassle to use.
> >
> > boost: no experience
> >
> > So I would personally be more inclined towards GSL, but only because I
> > had some experience with it.
>
> In fact, of all options I presented I also only have actual experience
> (beyond a few simple tests) with the GSL.
>
> > I have however heard that GSL is
> > not-so-perfect library ... so it may appear that eventually you want to
> > switch to boost - but for one function that should not be a big deal,
> right?
>
> Probably not. Can you say something more about GSL being not so perfect?
>
>

Nothing specific to tell - just vaguely remember that someone was
commenting about GSL this way. May well be not true or true for some aspect
:)



> >
> > Thinking of the future, boost may indeed be the best option, but I
> > would definitely do some testing on different platforms. If you give me
> > clear instructions, I am willing to do testing for Mac.
>
> Great!
>
>
> >
> > Yurii
> >
> > PS what is p.o.p. code?
>
> Proof of principle.
>
>
> Best,
>
> Lennart.
>
> >
> > On Fri, May 24, 2013 at 2:55 AM, L.C. Karssen <lennart at karssen.org
> > <mailto:lennart at karssen.org>> wrote:
> >
> >     Dear list,
> >
> >     One of the long standing feature requests for ProbABEL (#2086 in the
> >     tracker) has been adding p-values to the output files (at least for
> the
> >     most-often used additive model based on dosage files).
> >     Maarten and I have discussed this a while ago and we'd like to
> implement
> >     this (Maarten actually already wrote a proof of principle).
> >
> >     Before we go for a certain implementation I'd like to start a
> discussion
> >     on this list on how to deal with dependencies.
> >     For this specific example we have the following options:
> >     1) Implement p-value code ourselves (from scratch)
> >     2) Use the R math library
> >     3) Use the GNU Scientific Library
> >     4) Use the Boost library
> >
> >     Option 1 is a bad one, because writing good code for extreme
> p-values is
> >     not trivial and this problem has already been solved (by at least the
> >     three other options in the list).
> >
> >     As to options 2-4 I'd like to hear your opinion on the following:
> >     a) introducing dependencies on external libraries: with any of these
> >     solutions we introduce a dependency for the people running ProbABEL.
> >     Option 2 depends on R being present, 3 on the installation of the GSL
> >     and option 3 on Boost. Note that (for efficient matrix math we
> already
> >     recommend installing the Eigen headers, but that is only necessary
> when
> >     compiling, not when running).
> >     Personally I don't see this as a bad thing, especially since I can
> use
> >     the ./configure mechanism to disable the feature if the lib is not
> >     installed
> >     b) which lib would be best to use? Option 2 is easy (and we have
> >     Maarten's p.o.p. code based on that). Option 3 would have been my
> first
> >     choice and is also not difficult to program (also the GSL has been
> shown
> >     to have a stable implementation for low p-values [1]). Option 3 is
> more
> >     complex to program, but given the wide range of the Boost libs, we
> may
> >     be able to leverage some of its other features as well (e.g.
> >     reading/writing gzipped files).
> >
> >     So in short:
> >     ad 2) easy, but large dependency on R (will probably also work on
> >     Windows)
> >     ad 3) easy, simple, but may be more difficult for Windows (needs to
> be
> >     checked)
> >     ad 4) more difficult, but more possibilities for the future
> >
> >
> >     Any thoughts/suggestions?
> >
> >     Thanks,
> >
> >     Lennart.
> >     --
> >     -----------------------------------------------------------------
> >     L.C. Karssen
> >     Utrecht
> >     The Netherlands
> >
> >     lennart at karssen.org <mailto:lennart at karssen.org>
> >     http://blog.karssen.org
> >
> >     Stuur mij aub geen Word of Powerpoint bestanden!
> >     Zie http://www.gnu.org/philosophy/no-word-attachments.nl.html
> >     ------------------------------------------------------------------
> >
> >
> >     _______________________________________________
> >     genabel-devel mailing list
> >     genabel-devel at lists.r-forge.r-project.org
> >     <mailto:genabel-devel at lists.r-forge.r-project.org>
> >
> https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/genabel-devel
> >
> >
> >
> >
> > --
> > -----------------------------------------------------
> > Yurii S. Aulchenko
> >
> > [ LinkedIn <http://nl.linkedin.com/in/yuriiaulchenko> ] [ Twitter
> > <http://twitter.com/YuriiAulchenko> ] [ Blog
> > <http://yurii-aulchenko.blogspot.nl/> ]
>
> --
> -----------------------------------------------------------------
> L.C. Karssen
> Utrecht
> The Netherlands
>
> lennart at karssen.org
> http://blog.karssen.org
>
> Stuur mij aub geen Word of Powerpoint bestanden!
> Zie http://www.gnu.org/philosophy/no-word-attachments.nl.html
> ------------------------------------------------------------------
>
>


-- 
-----------------------------------------------------
Yurii S. Aulchenko

[ LinkedIn <http://nl.linkedin.com/in/yuriiaulchenko> ] [
Twitter<http://twitter.com/YuriiAulchenko>] [
Blog <http://yurii-aulchenko.blogspot.nl/> ]
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.r-forge.r-project.org/pipermail/genabel-devel/attachments/20130526/5fbe7b63/attachment.html>


More information about the genabel-devel mailing list