← Back to team overview

ubuntu-scientists team mailing list archive

Re: Idea for a Collaboration Project

 

Regarding reproducible research, I have adopted the strategy of creating R
packages for every analysis I do.
I have defined a small R snippet in my IDE (Architect
http://www.openanalytics.eu/downloads/architect) that creates a new R
package based on the selected project in Eclipse (the basis of Architect).
Then I just add Sweave (or knitr) files for the reports in /vignettes, and
data files go into the package as well. So all code, reports and data are
all together in one package.
Dependencies are then just declared on the package level, and will even be
automatically installed if you install the package.
(In the company I work for we even have an internal R repository where
these packages can go in, and are then available for all other employees to
install.)

Maybe I should take some time to write this all down, and maybe write a
small R package to make some of the steps a bit easier.
But the thing is, that most people don't change the way they work, so I am
not sure how much impact it would have.

Kind regards,

Willem

On Thu, Oct 30, 2014 at 1:58 PM, Sheila Miguez <sheila.miguez@xxxxxxxxxxxxx>
wrote:

> Yes, I am impressed at the R community.
>
> There is http://cran.r-project.org/web/views/ReproducibleResearch.html
>
> and rOpenSci http://ropensci.github.io/ which has spawned pyOpenSci
> https://github.com/pyOpenSci
>
> There are a lot of enthusiastic people (many of them involved in Software
> Carpentry) who are participating in a lot of initiatives...
>
>
> Mozilla Science Lab has been doing a lot, in
> http://forum.mozillascience.org/t/code-review-project/22/ I suggested
> maybe trying to work on a cookiecutter project for different types of
> research.
>
> For doing reproducible research, trying to capture dependencies is very
> tricky, so anything that helps with that would be valuable. People who are
> pros at package management could certainly help there.
>
> People who are pros at containers and virtual machines will also have
> valuable insights.
>
> There are custom distros for scientists, like neuro debian. I am more
> enthusiastic about packages over custom distros.
>
>
> On Thu, Oct 30, 2014 at 4:34 AM, Willem Ligtenberg <wligtenberg@xxxxxxxxx>
> wrote:
>
>> Hi all,
>>
>> Just to pitch in on the R part, since that is my domain.
>> Specifically for R a lot of code is shared through packages, either
>> through CRAN, Bioconductor or sometimes githubs.
>> Obviously, some people still keep code for themselves, but specifically R
>> is doing a pretty good job at making code sharing easy.
>>
>> One initiative that I am aware of is:
>> http://software-carpentry.org/index.html
>>
>> Kind regards,
>>
>> Willem
>>
>> On Thu, Oct 30, 2014 at 7:08 AM, A. Mani <a.mani.cms@xxxxxxxxx> wrote:
>>
>>> On Wed, Oct 29, 2014 at 5:18 AM, Walter Lapchynski <wxl@xxxxxxxxxx>
>>> wrote:
>>> > Sounds interesting, but I'm not sure I understand what you have in
>>> mind.
>>>
>>>
>>> I meant we should have people exploring domain specific practices and
>>> requirements (within the sciences).
>>>
>>> For example, a few days ago I posted a mail on aspects relating to
>>> LaTeX- related work flows.
>>>
>>> I think there are many other work flows that can be the reason for
>>> further customization.
>>>
>>> Consider the case of people doing machine learning on GNU/R.
>>>
>>> The relevant "views" is not the last word.
>>>
>>> People write their own code variants, but do not care to publish.
>>>
>>> There are far too many ways of implementing algorithms from scratch or
>>> at some mixed level.
>>>
>>> We can have projects to help with the situation.
>>>
>>>
>>> specify ...there are thousands more.
>>>
>>>
>>>
>>> Best
>>>
>>> A. Mani
>>>
>>>
>>>
>>> A. Mani
>>> [Last_Name. First_Name Format]
>>> CU, ASL, AMS, ISRS, CLC, CMS
>>> HomePage: http://www.logicamani.in
>>> Blog: http://logicamani.blogspot.in/
>>> sip:girlprofessor@xxxxxxxxx
>>>
>>> --
>>> Mailing list: https://launchpad.net/~ubuntu-scientists
>>> Post to     : ubuntu-scientists@xxxxxxxxxxxxxxxxxxx
>>> Unsubscribe : https://launchpad.net/~ubuntu-scientists
>>> More help   : https://help.launchpad.net/ListHelp
>>>
>>
>>
>> --
>> Mailing list: https://launchpad.net/~ubuntu-scientists
>> Post to     : ubuntu-scientists@xxxxxxxxxxxxxxxxxxx
>> Unsubscribe : https://launchpad.net/~ubuntu-scientists
>> More help   : https://help.launchpad.net/ListHelp
>>
>>
>
>
> --
> sheila.miguez@xxxxxxxxxxxxx
>

References