[R-wiki] [Fwd: The results of your email commands]

[Frank E Harrell Jr]

Philippe Grosjean wrote:
> Frank E Harrell Jr wrote:
> 
>> Philippe Grosjean wrote:
>>
>>> Frank E Harrell Jr wrote:
>>>
>>>> [...]
>>>>
>>>> I think that's enough, just think that Data Manipulation should be 
>>>> promoted to a section.
>>>>
>>>> Frank
>>>
>>>
>>>
>>>
>>> One has to decide and keep consistent: the current top level (level 
>>> 1) division of sections is between short docs (going to 'snippets' 
>>> section) and long docs (going to 'tutorials' section). Then, we have:
>>> 1) in 'snippets': 'data-manip' among others.
>>> 2) in 'tutorial': no equivalent because longer document are likely to 
>>> discuss various things... not just data manipulation. If, by chance, 
>>> there is a longer document that discusses *only* data manipulation, 
>>> it is possible to add a 'data-manip' subsection in 'tutorial' too.
>>>
>>> If we make 'data-manip' a top level section, we break the whole 
>>> logic, and there is no reasons we shoudn't promote 'stats', or 
>>> 'graphics' for instance as top-level sections... and at the end, we 
>>> rework everything and end up with a totally different structure.
>>>
>>> Just think at the structure on harddisk in a more practical way. I 
>>> like to separate short and long documents, because they are managed 
>>> very differently. Long documents have a primary author that is more 
>>> susceptible to update his own work, to append to it, etc... Many sort 
>>> documents will be one shot (typically, someone converts an 
>>> interesting R-Help thread, for instance). It is really more a 
>>> collaboration of many people, it is much more chaotic. Consequently, 
>>> it requires a larger number of subsections to keep a little bit of 
>>> order. I think, for these reasons, that we have now a pretty good 
>>> structure to help manage the Wiki.
>>>
>>> Now, it is possible to write tables of content that are more 
>>> task-specific, that is, promote 'data', 'stats', 'graphics' as top 
>>> level, and list all documents (both from 'tutorials' and 'snippets') 
>>> under these sections. One could also think about such table of 
>>> contents  as area-specific (biology, finance, etc...). That is what 
>>> is experimented here: 
>>> http://www.sciviews.org/_rgui/wiki/doku.php?id=tutorials:tutorials
>>>
>>> Does this answers to your request (without breaking the current 
>>> structure)?
>>
>>
>>
>> Philippe, you have a better understanding of this than I do, so I'm 
>> comfortable with your approach.  I don't recommend area-specific main 
>> classifications as their is much overlap between areas in the stat and 
>> graphical methods used (and especially in data manipulation).
>>
>> Frank
> 
> 
> Do you mean it is not a good idea to separate 'data' from 'stats' from 
> 'graphics'? Gosh! It is the clearest separation we could do in my mind 
> (even if there are overlaps). What do you porpose? Please, keep in mind 
> that it is just a convenient way to have a little bit of order in the 
> files *on the server's disk* (that is, you have to decide to place a 
> file in one category on the disk, but you can refer to it from different 
> places in the table of contents).

No, I was referring to your comment biology, finance, etc... - I 
wouldn't separate things by area of application.

Frank

> 
> One example: a page deals with residuals analysis of regressions. So, is 
> it 'stats-models' or 'graphics-base', given that there is a wide 
> discussion on the graphs also? Well, it doesn't matter! Just choose one 
> (let's say 'stats-models'), and refer to the page from both 
> 'stats-models' and 'graphics-base' in the table of contents.
> Best,
> 
> Philippe
> 
> 


-- 
Frank E Harrell Jr   Professor and Chair           School of Medicine
                      Department of Biostatistics   Vanderbilt University