R help - Oct 2010 - R: Tools for thinking about data analysis and graphics

I'm giving a talk about some aspects of language and conceptual tools 
for thinking about how
to solve problems in several programming languages for statistical 
computing and graphics. I'm particularly
interested in language features that relate to:

o expressive power: ease of translating what you want to do into the 
results you want
o elegance: how well does the code provide a simple human-readable 
description of what is done?
o extensibility: ease of generalizing a method to wider scope
o learnability: your learning curve (rate, asymptote)

For R, some things to cite are (a) data and function objects, (b) 
object-oriented methods (S3 & S4); (c) function mapping over data with 
*apply methods and plyr.

What other language features of R should be on this list?  I would 
welcome suggestions (and brief illustrative examples).

-Michael


-- 
Michael Friendly     Email: friendly AT yorku DOT ca
Professor, Psychology Dept.
York University      Voice: 416 736-5115 x66249 Fax: 416 736-5814
4700 Keele Street    Web:   http://www.datavis.ca
Toronto, ONT  M3J 1P3 CANADA

Hi,

On Wed, Oct 6, 2010 at 5:05 PM, Michael Friendly <friendly at yorku.ca>
wrote:
> ?I'm giving a talk about some aspects of language and conceptual tools
for
> thinking about how
> to solve problems in several programming languages for statistical
computing
> and graphics.
For graphics, I'm guessing you'd mention something about ggplot's
grammar of graphics.

Also, although I don't use it all that much, I think the "formula"
methods (y ~ a + b)-type of stuff is probably handy for certain
audiences, as it somehow translates to a "natural language" when doing
modeling.

-- 
Steve Lianoglou
Graduate Student: Computational Systems Biology
?| Memorial Sloan-Kettering Cancer Center
?| Weill Medical College of Cornell University
Contact Info: http://cbio.mskcc.org/~lianos/contact

On 10/6/2010 6:14 PM, David Winsemius wrote:
>
> Graphics: I realize you were focusing on "language" but the
graphical
> tools are extremely important if one is describing how data 
> exploration and summarization is done effectively.
Well, I'm focusing on the language features that make it easy or hard to 
draw a standard sort of graph
(controlling details) or to design a new kind of graph.  Base graphics, 
grid + lattice, ggplot all have
strengths (and weaknesses) I'm thinking about.

Replying to Bert Gunter:  Yes, I agree that the formula language for 
lm/glm and friends,
plus extensions to other aspects (plots, tables, etc.) is another 
important R feature of
the type I'm considering.  I would put this in the top 5!

-Michael

-- 
Michael Friendly     Email: friendly at yorku.ca
Professor, Psychology Dept.
York University      Voice: 416 736-5115 x66249 Fax: 416 736-5814
4700 Keele Street    http://www.math.yorku.ca/SCS/friendly.html
Toronto, ONT  M3J 1P3 CANADA

On Wed, Oct 6, 2010 at 4:05 PM, Michael Friendly <friendly at yorku.ca>
wrote:
> ?I'm giving a talk about some aspects of language and conceptual tools
for
> thinking about how
> to solve problems in several programming languages for statistical
computing
> and graphics. I'm particularly
> interested in language features that relate to:
>
> o expressive power: ease of translating what you want to do into the
results
> you want
> o elegance: how well does the code provide a simple human-readable
> description of what is done?
> o extensibility: ease of generalizing a method to wider scope
> o learnability: your learning curve (rate, asymptote)
>
> For R, some things to cite are (a) data and function objects, (b)
> object-oriented methods (S3 & S4); (c) function mapping over data with
> *apply methods and plyr.
>
> What other language features of R should be on this list? ?I would welcome
> suggestions (and brief illustrative examples).
 * missing values
 * subsetting
 * lexical scope and closures (goes along with first class functions)
 * built-in documentation
 * CRAN (not exactly a language feature, but important part of ecosystem)
 * thoughtful interactive features - e.g. a <- 10 doesn't print 10.


Hadley

-- 
Assistant Professor / Dobelman Family Junior Chair
Department of Statistics / Rice University
http://had.co.nz/

Hi, Michael,

When I teach/preach on R, I emphasize the language's focus on data,
both in its objects and operations. It might seems basic, but it's
fundamental to most of the features you and others have mentioned. As
a statistical programming language, what we intend to do with R is
often very naturally accomplished using vector operations on tabular
data, where columns represent variables of the same data type and rows
represent observations of these variables for a given member of the
dataset.  Fortunately, these are core components of R.  For instance,
we can easily perform complex selections of variables and/or members,
which, more often than not, serve as input to or power the functions
that generate the statistics and graphics we care about.
Unfortunately, vector operations seem to be difficult for people to
learn how to use properly, and there are penalties for not using them,
but as they say: no pain, no gain. :)

If you'd be willing to share the materials you create for your talk,
I'd be interested in seeing them.

Cheers,

Jeff.

On Wed, Oct 6, 2010 at 5:05 PM, Michael Friendly <friendly at yorku.ca>
wrote:
> ?I'm giving a talk about some aspects of language and conceptual tools
for
> thinking about how
> to solve problems in several programming languages for statistical
computing
> and graphics. I'm particularly
> interested in language features that relate to:
>
> o expressive power: ease of translating what you want to do into the
results
> you want
> o elegance: how well does the code provide a simple human-readable
> description of what is done?
> o extensibility: ease of generalizing a method to wider scope
> o learnability: your learning curve (rate, asymptote)
>
> For R, some things to cite are (a) data and function objects, (b)
> object-oriented methods (S3 & S4); (c) function mapping over data with
> *apply methods and plyr.
>
> What other language features of R should be on this list? ?I would welcome
> suggestions (and brief illustrative examples).
>
> -Michael
>
>
> --
> Michael Friendly ? ? Email: friendly AT yorku DOT ca
> Professor, Psychology Dept.
> York University ? ? ?Voice: 416 736-5115 x66249 Fax: 416 736-5814
> 4700 Keele Street ? ?Web: ? http://www.datavis.ca
> Toronto, ONT ?M3J 1P3 CANADA
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>

I think that R/S's biggest advantage is in the ways you can store data.  It
does not force you to fit your data to a single spreadsheet like table, but
rather encourages you to think about your data and find the correct way to store
it.  Lists and objects are a great advantage for keeping related things combined
together.  I can have multiple data sets available all at the same time, but
still in separate objects.

Also the results of routines can be kept in a way that makes working with them
easy.  I remember working with programs that just had one big spreadsheet and
ending up with 3 different columns of residuals from 3 different models, but
then forgot which residuals matched which model.  With R/S each lm object has
the residuals stored with it including the call to remind us what model and
options were used.  One plot that I like to make when exploring different models
is:
> plot( fitted(model1), fitted(model2) ); abline(0,1)
That is simple and straight forward in R/S, but much more difficult in other
programs.


I also like the fact that the graphics system will let me create anything I
want.  There are tools to create the standard plots (and I really like the
simplicity of calling plot on an lm object and getting a standard set of
diagnostics), but there are also the tools to create any plot I can imagine, or
add any information I feel useful to an existing plot.

-- 
Gregory (Greg) L. Snow Ph.D.
Statistical Data Center
Intermountain Healthcare
greg.snow at imail.org
801.408.8111

> -----Original Message-----
> From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-
> project.org] On Behalf Of Michael Friendly
> Sent: Wednesday, October 06, 2010 3:06 PM
> To: R-help
> Subject: [R] R: Tools for thinking about data analysis and graphics
> 
>   I'm giving a talk about some aspects of language and conceptual tools
> for thinking about how
> to solve problems in several programming languages for statistical
> computing and graphics. I'm particularly
> interested in language features that relate to:
> 
> o expressive power: ease of translating what you want to do into the
> results you want
> o elegance: how well does the code provide a simple human-readable
> description of what is done?
> o extensibility: ease of generalizing a method to wider scope
> o learnability: your learning curve (rate, asymptote)
> 
> For R, some things to cite are (a) data and function objects, (b)
> object-oriented methods (S3 & S4); (c) function mapping over data with
> *apply methods and plyr.
> 
> What other language features of R should be on this list?  I would
> welcome suggestions (and brief illustrative examples).
> 
> -Michael
> 
> 
> --
> Michael Friendly     Email: friendly AT yorku DOT ca
> Professor, Psychology Dept.
> York University      Voice: 416 736-5115 x66249 Fax: 416 736-5814
> 4700 Keele Street    Web:   http://www.datavis.ca
> Toronto, ONT  M3J 1P3 CANADA
> 
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-
> guide.html
> and provide commented, minimal, self-contained, reproducible code.