[R] Another NEWBIE

[Frank E Harrell Jr]

Charles and Kimberly Maner wrote:
> Hi Frank.  I am (somewhat) new to R as well, but almost a 10 yr SAS veteran.
> I work for a very large US Bank and have spent a considerable part of my
> career in Corp Mktg leveraging data for, arguably, data mining, next
> purchase, attrition, balance diminishment and the like.  I am now managing
> an Operations Research group in their Customer Service and Support (aka
> Telephone/Call Center Support) within the forecasting and analytics group.
> What I have found, broadly and personally, regarding R vs. sAS is the
> following:
> 
> 1.  You simply can't beat the price of R vs. Insightful Corp.'s S-Splus, not
> to mention SAS.
> 2.  The support folks for R are among the very best, (e.g., most helpful,
> energetic and enthusiastic to help)
> 3.  R is far, far leaner from what I have seen thus far for modeling,
> binning/discretizing, graphing, etc. vs. SAS.

Thanks for your note.  I assume you meant to say 'more capable' rather 
than 'leaner'.

> 4.  SAS is, per a previous post, (quite debatably) superior for manipulation
> and handling of fantastically large datasets.  I have found that R's
> strength is not really in merging datasets and dataset manipulation.
> Although, major caveat here, it greatly depends on what you need done to the
> data.  For lagging, diffing, binning, R is superior.  For match merging, at
> this stage, I vote for SAS.  (Again, I stress I've only 6-8 mos of moderate
> R experience.)

You are right, for huge datasets.  For others, R is great, even for 
merging.  Many examples are provided in the Alzola and Harrell text on 
http://biostat.mc.vanderbilt.edu

> 5.  The challenge with R is, perhaps, it's very strength--language density.
> Once I learn how to do something in R vs. SAS, R's code is fractionally as
> large as SAS.  Literally, it may take 10 lines of code in SAS vs. a one
> liner in R.  That's powerful.  However, due to my SAS experience, I've
> banged out the SAS code and am still looking/hunting for the R equivalent.
> However, once doing so, it's, borrowing from a popular vegetable drink
> slogan, "Wow, I could have done that in R."

Yes I agree.

> 6.  And, lastly, while R is well documented, I seem to find one of the areas
> of documentation somewhat lacking is a great big R "recipe" book.
> (Suggestions, BTW, are welcome here.)  Documentation of the R language is in
> place with more being published, (alongside S-Plus), annually.  However,
> there does not appear to be, for example, an "R Transition Recipes for
> Experienced SAS Users" book.  That, ultimately, is what would help me, (I
> think.)  Again, the issue really is simply learning and using the language.
> Experienced R users, I'm convinced, could do everything R I'm doing in SAS,
> (with money left over for a few coffees at Starbuck's).

I agree.  What I really think is needed is a compendium of examples, 
especially for data manipulation.  I gave a talk about this last week; 
abstract is at 
http://biostat.mc.vanderbilt.edu/twiki/bin/view/Main/FrankHarrellrmanip 
with links to other places.  A meager attempt at navigating some of the 
more commonly used R functions is at 
http://biostat.mc.vanderbilt.edu/s/finder/finder.html

> 
> In conclusion, I still think that, given one's budget and projects, there's
> a place for SAS and R to co-exist.  But, that paradigm diminishes as (1) the
> size of the datasets become smaller and, (2) your problems are more
> academic/researchy/specific in nature.  For graphing, esp. w/the Lattice
> package, R is simply superior (IMHO), period, to SAS.  (For some reason, SAS
> has just not felt the need to improve their graphics, at least the SAS/Graph
> part of their offering.)  And, for the SAS lovers out there, this opinion is
> mine only as I continue to be primarily a SAS client attempting to
> transition to R.
> 
> Frank, while I've probably been too wordy, I've attempted to provide another
> perspective for you.  Good luck.

No, well said,

Thanks,

Frank

> 
> 
> Thanks,
> Charles
> 
> 
> ------------------------------
> 
> Message: 7
> Date: Sat, 19 Jun 2004 18:15:19 +0200 (MEST)
> From: "F.Kalder" <Kalderf at gmx.de>
> Subject: Re: [R] Another NEWBIE
> To: r-help at stat.math.ethz.ch
> Message-ID: <6411.1087661719 at www45.gmx.net>
> Content-Type: text/plain; charset="us-ascii"
> 
> Hi,
> 
> Thank you all who anwered me. 
> 
> I think, I mainly thought to understand the difference between SPSS /SAS and
> R, but didn't really get the point (what explains the question, wich metods
> R can't do). Maybe, because I don't have much experience with programming
> (near to none). My background in stats goes also only back to indroductory
> classes and an advanced course in multivariate statistics. To this, I'm
> working with Hair, Anderson, Tatham & Blacks's "Multivariate Data Analysis"
> (5th Ed.) as my ressource, mainly with questionnaire analysis (Reliability
> Analysis and Factor Analysis, also MDS, Conjoint etc. plus sometimes
> standard MANOVA, Multiplke Regression etc.). So, maybe my stats aren't
> sophisticated enough to use R, I'm just a standard user of applied
> statistical methods, not an academic researcher or even a statistician. It
> was mainly a descision by costs, because R is free software. 
> With the concept, I completely mistook the R concept as a programming
> environment more as a kind of advanced SPSS Syntax (because I also would
> call it "programming" when using it), which I now know, is completely wrong.
> 
> So, I again thank for your help.
> 
> 
> Cheers, Frank.
> 
> --



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