[R] functions different in R and S

[Spencer Graves]

Also, many of the probability distribution functions are better in R
1.6.2 than in S-Plus 6.1.  For example, in R, most of the functions will
work with the logarithm of the probability (or density) using argument
"log.x";  S does not have this argument.Simularly, probability functions 
that could have a noncentrality parameter are more likely to have it in 
R than in S-Plus.

I had trouble over a year ago with numerical precision with some 
probability functions in S-Plus 2000, but I can't recall the details 
now.  If you are estimating degrees of freedom in Student's t 
distribution, e.g., to model kurtosis, you don't want the functions to 
malfunction for unusual values of the argument df.  I just sent S-Plus 
6.1 into an incredibly long computation with pt(-1.96, df=1e99);  R 
returned a sensible answer immediately.  If you are trying to estimate 
df numerically, it is better not to have this.

Spencer Graves

David Brahm wrote:
 > Helen Yang <Huiqin.Yang at noaa.gov> wrote:
 >
 >>I wonder whether there is any resource that can point to useful 
substitutes
 >>for S functions that are not recognized by R.  At the same time 
whether there
 >>is a list of functions, which appear in both R and S but which don't do
 >>exactly the same thing.
 >
 >
 > As Prof. Brian D. Ripley <ripley at stats.ox.ac.uk> said, see the FAQ at:
 >   <http://www.r-project.org/> (Click "FAQs", "R FAQ", "R and S".)
 >
 > Here are my own crib notes on some differences that I care about. 
Note "S"
 > here really means S-Plus version 6.1.2.  I have not checked whether 
any of
 > these have changed since October.  This list is certainly not complete.
 >
 >                      ***   R vs. S (DB 10/28/02)  ***
 >
 > Language differences:
 > - Scoping rules differ.  In R, functions see the functions they're 
in.  Try:
 >     f1 <- function() {x <- 1; f2 <- function() print(x); f2()};  f1()
 > - Data must be loaded explicitly in R, can be attach()'ed in S.
 >     Addressed by my contributed package "g.data".
 > - R has a character-type NA, so LETTERS[c(NA,2)] = c(NA,"B") not 
c("","B")
 > - paste("a","b", sep="|", sep=".") is an error in R; ok in S.
 > - for() loops more efficient in R.
 >
 > Graphics differences:
 > - Log scale indicated in S with par(xaxt)=="l", in R with par("xlog")==T.
 > - R has cex.main, col.lab, font.axis, etc.  Thus title("Hi", cex=4) 
fails.
 > - R has plotmath and Hershey vector fonts.
 > - R has palette(rainbow(10)) to define colors (both screen and printer).
 >
 > Functions missing from R:
 > - unpaste, slice.index, colVars
 >
 > Functions missing from S:
 > - strsplit, sub, gsub, chartr, formatC
 >
 > Functions that work differently:
 > - system() has no "input" argument in R.
 > - substring(s,"x") <- "X" only works in S, but R has s <- 
gsub("x","X",s).
 > - scan expects numbers by default in R.
 > - which(<numeric>) converts to logical in S, is an error in R.
 > - The NULL returned by if(F){...} is invisible in R, visible in S.
 > - The NULL returned by return() is visible in R, invisible in S.
 > - Args to "var" differ, and R has "cov".  S na.method="a" ~ R use="p".
 > - var (or cov) drops dimensions in S, not R.
 > - cut allows labels=F in R, not in S (also left.include=T becomes 
right=F).
 > - Last argument of a replacement function must be named "value" in R.
 > - tapply(1:3, c("a","b","a"), sum) is a 1D-array in R, a vector in S.
 >