[R] Teaching R in 40 minutes. What should be included?

[Frank E Harrell Jr]

Spencer Graves wrote:
>      I agree with Thomas and Georg:  A 40 minute intro should be mostly 
> Marketing and very little "how to".
>      I think you'll have a more effective sales job if you target, say, 
> 4 examples averaging 5 slides each plus some general overview, max 25-30 
> slides.  If I had sufficient prep time and a few collaborators among 
> physicists, geographers, etc., I might get their help in preparing 
> examples, showing how they would do something in Matlab or Scilab or 
> something else vs. R.  And I'd end with a discussion of technical 
> support via an R site search and r-help and showing a list of available 
> contributed packages.  I'd do a couple of searches for physics and 
> geographical questions.  ODESOLVE, maps, etc.  Maybe pick examples that 
> are part of the help files.  Then show, here is how I find X, here is 
> the vignette, help or whatever.
>      Is it fair to say that R is rapidly becoming (if it is not already) 
> the primary platform of choice for new statistical algorithm 
> development?  I think they might be interested in a brief overview of 
> the contributed software.  If this is an academic audience, they might 
> like to know how easy it is to contribute software, plus journal on 
> statistical computing and graphics, etc.
>      hope this helps.  Good Luck!
>      spencer graves   

I often give talks like that.  The thing that has impressed audiences 
the most is a multi-panel lattice graphic with 2 classification 
variables and in each panel a scatterplot and a lowess trend line.  A 
single page with 24 small high-resolution histograms also seems to 
impress people.  The nomogram function in the Design package seems to 
also connect with non-statisticians, as does 
latex(describe(mydataframe)) using Hmisc.  People like seeing in the 
latex previewer some output that mixes tabular summaries and graphics.

Frank Harrell

> Thomas Schönhoff wrote:
> 
>> Hello,
>>
>> Am Freitag, 25. Februar 2005 22:37 schrieb Dr Carbon:
>>  
>>
>>> If _you_ were asked to give a 40 minute dog and pony show about R
>>> for a group of scientists ranging from physicists to geographers
>>> what would you put in? These people want to know what R can do.
>>>
>>> I'm thinking about something like:
>>>
>>> A. Overview
>>> B. data structures
>>> C. arithmetic and manipulation
>>> D. reading data
>>> E. linear models using glm
>>> F. graphics
>>> G. programming
>>> H. other tricks like rpart or time series analysis?
>>>   
>>
>>
>> If your audience is well known I would be inclined to target some 
>> (simple) examples derived from physics and geography to demonstrate 
>> basic ideas of working with R, similar like the ones listed above.
>>
>> Well, 40 minutes are not too long, so I recommend to simplify your 
>> presentation as much as you can. You want teach them R in 40 minutes 
>> but rather tend to confuse them if you don't shorten your plan a bit.
>> I.E. teaching programming in R in a few minutes for scientists who are 
>> not at all acustomed to programming  is much overhead, I think.
>> Well, it's up to your estimation on what is expected to follow your 
>> presentation. If you are sure that most of them know enough 
>> programming to unterstand the basic concepts in R-programming, 
>> everything will be fine!
>> If not, I'd recommend  to concentrate on basic operations (data 
>> structures, arithmetic and manipulation, import/export data and some 
>> often used default statistical procedures demonstrating common tasks 
>> (is time series analysis important in physics or geography, I don't 
>> know??), including some remarks on diffenrences to widespread 
>> statistical packages like SPSS or SAS, maybe LispStat.
>> Finally there shouuld be some extended view of available ressources 
>> (manuals, FAQ, community) as a starter to learn, use and program R by 
>> themselves.
>> I think this would do for a 40 minutes presentation without taking the 
>> risk to deter people due to overcomplexity.
>>
>> regards
>> Thomas

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