[R] R Graph Gallery : categorization of the graphs

[Sander Oom]

I agree that a wiki to facilitate submission of graph code could be very 
effective! Still needs to be well protected against vandalism. Seems a 
regular backup, to facilitate a clean restore, is the best approach.

Romain, would you be willing to set up a wiki within the gallery. Think 
the wiki and gallery should be close to each other in cyber space!

I use DokuWiki privately and it works wonders.

Guess the database design will require a bit of thought. For each graph 
it should be possible to enter multiple categories and sub-categories. 
then the gallery interface should dynamically build galleries using 
these categories. Not sure if you would want to create 'meta' categories 
to facilitate multiple categorization approaches!?

We could start with a limited list such as suggested by Chris to 
populate the gallery in the first instance.

Pleased to see this progress!!

Sander.


Chris Evans wrote:
> On 6 Jun 2005 at 17:48, Sander Oom wrote:
> ... much snipped ...
>>The whole point of a gallery is to show something to the user before
>>the user knows what he is looking for. The R help functions currently
>>available are hopeless when you have a picture of a graph in your head
>>without knowing the required commands.
> ... much snipped ...
> 
> Belief that good graphics are often as important or more important 
> than inferential tests or even CIs was one of the reasons I've moved 
> over the last 15 years from SPSS (still use it a bit 'cos most 
> colleagues do) through SAS (much better, much better graphics) to S+ 
> (same but more so) to R (same and FLOSS!)
> 
> The demo graphics and the gallery are wonderful visual arguments for 
> R and also great resources to help us learn.  Categories that I think 
> might be useful sometimes might be: 
>   	describes one variable where that is:
> 		dichotomous, categorical (n(categories) > 2), polytomous (short),
> 			polytomous (many levels), or continuous (might allow something on
> 			superimposing different referential distributions
> 	describes relationship between two variables where:
> 		both are dichotomous or polytomous
> 		one is ditto, other is continuous (box & violin etc: very useful to
> 			see good e.g.s of how to get most appropriate boxplots as it's 
> 			always possible to get good ones but not always obvious)
> 			(pointer to back-to-back histogram in Hmisc here)
> 		both are continuous (with and without jitter and weighting blobs)
> 	describe relationships between more than two variables...
> 
> However, the gallery idea is a very powerful one and being able to 
> scroll through and drill down is a useful trick that M$ have, I 
> grudgingly admit, used well so could we mimic their galleries from 
> Excel as someone has suggested and perhaps mimic the drop down 
> graphics picker in S+ (I no longer have access).
> 
> It's not much help but someone could put up the drop down list for 
> newbies coming from SPSS ... ooh, just opened up my copy (11.0.1) and 
> realised there's a gallery there with the following:
> bar, line, area, pie, high-low, pareto, control, boxplots, error bar, 
> scatter, histogram, normal P-P, normal Q-Q, sequence, 
> autocorrelations, cross-correlations & spectral.  Never knew that the 
> was there!  The drop down list below that has essentially the same 
> list but with the last three under a sub-heading of "time series" and 
> ROC curves added.
> 
> I wonder if someone hosted a wiki for a while at least it would get 
> people contributing code for examples for some of these?  The results 
> could transfer to the wonderful graphics gallery as they accumulated. 
>  My skills aren't that hot but I'd throw in a few things happily and 
> I'm sure a reward would be hearing of better ways to do things both 
> in terms of coding and in terms of better displays/graphics to use.
> 
> Cheers all,
> 
> C


-- 
--------------------------------------------
Dr Sander P. Oom
Animal, Plant and Environmental Sciences,
University of the Witwatersrand
Private Bag 3, Wits 2050, South Africa
Tel (work)      +27 (0)11 717 64 04
Tel (home)      +27 (0)18 297 44 51
Fax             +27 (0)18 299 24 64
Email   sander at oomvanlieshout.net
Web     www.oomvanlieshout.net/sander