[R] Comparing R and SAS

[Terry Therneau]

Satish,
  There are nearly as many opinions as people on this question.  So accept these 
as my views only.
  
  1. I have written S packages, SAS procedures and SAS macros over my career.  S 
was specifically designed for extensibility and it shows.  The ratio of time to 
get a new statistical idea up and running is about 1:5:20 for S:macro:proc, 
using the "therneau days" metric.  New ideas are the lifeblood for an academic; 
SAS makes no sense in that environment and so will always lag far behind.  For 
simple things like a repetive recoding, a 2-10 line R function say, SAS macro 
will be competive with R in terms of programming effort.  
  
  2. The SAS programming model could be described as "mini batch"; the block of 
statements including and following a "data" statment is executed as a unit.  For 
data manipulation this is turns out to be a very powerful programming paradym 
and many people, myself included, prefer it for data input and manipulation.  At 
the time I was first using SAS (mid 70s) no other package or system was even 
close to SAS in this arena, and by the late 80s many disciplines with complex, 
large, or messy data had wedded themselves to SAS, e.g., drug company clinical 
trials.  Most are still wedded.  
  
  Terry Therneau