Dear S-Users,
Thanks to Richard Boylan, Patrick Connoly, Derek Eder, Jim Garrett, Raphael
Gottardo, Dave Kane, Pierre Kleiber and Andy Liaw for their comments
regarding switching from Splus to R. I have synthesized their comments
under various headings. I assume responsibility for any inadvertent
misrepresentation of their thoughts.
Most people suggested that the switch depends on usage. Some features are
available in one package that are not available in the other. R might be
faster, with a smaller footprint and may be more bulletproof when reading
data. The degree of support was mentioned by some respondents. Splus
offers technical support while R users rely on a newsgroup (in my opinion a
very helpful group). The general consensus is that the packages are
similar.
My Splus usage is a single machine in a consulting environment. I find the
GUI irritating, and do not require the connectivity, Web applets etc. My
impression of Insightful's move toward a corporate computing environment is
that I am paying for a large degree of overhead I do not want. If Insightful
had not added the two new libraries I would not consider the downgrade at
all. However, I did switch to Splus from SAS due to the flexibility of its
programming language. The programming language and command line interface
is still there.
I think I will downgrade to the new version, but also try out R.
Sincerely,
Barry Zajdlik
Splus vs R Features
The following is a non-exhaustive list of features pointed out by the
various respondents.
Splus Trellis graphics more advanced than R Lattice
Tech support from Splus but possibly not for legacy software, user group for
R
Multicomp in Splus, not in R
Hairline Plots in Splus, not in R
Splus can import and export data sets in other formats
Splus saves variables automatically.
Control charting available in Splus, not R
R's GAM package allows for automatic selection of smoothing constants (and
it does a very good job of
it!) while this is not implemented in S-Plus's
If you apply a Chi-squared test for independence, you can opt to estimate
the exact p-value by simulation, which again works very well and is not
implemented in S-Plus(at least, it's not integrated into the Chi-squared
test function)
In addition, you will probably find R to be a bit faster than S-Plus
overall, easier to maintain, smaller footprint--all around a pleasant
surprise. Not radically better, but what you would expect for a product
that gives very good programmers a chance to improve on a very good
original.
New Splus 6 features
For more on this, Insightful offers a 1/2 hour web presentation.
missing and robust libraries,
improved integration with Excel,
new and improved graphics export to Powerpoint,
many bug fixes
Comments on the New Class System
One more thing: In case you're not aware, John Chamber has been working on
the new class system (as in Splus 6) for R. It is available now as an R
package. I was told by Prof. Lumley (one of the R Core developers) that it
will be in the base distribution of R (although its use is optional). It is
entirely possible that R will also become Sv4 in version 2.0. At least
there seems to be pretty strong intension among R Core members (John
Chamber, Brian Ripley, Luke Tierney, etc.) to move in that direction.
The bottom line is, if you are worrying about the issues Dr. Thernau and
Prof. Harrell raised about the new class system, moving to R may only be of
temporary relief.
For most users, the portability issue is minor to non-existence, I suspect.
We have some code that uses the class mechanism in Splus 2000. I was able
to use these code in Splus 6 on Linux (available almost a year ago) almost
w/o modification (the modifications are needed for the difference between
platforms, not Splus versions). I may get a couple of warning messages, but
they are quite harmless. I didn't use the migration tools at all. Andy Liaw
I've stayed away from Splus 6.0 (and 5) because I didn't want to redo all
that new class stuff. Maybe it would not have been any more work than
porting code across to R (the scoping rules make some things very
different -- some easier, some more difficult). Patrick Connolly
R-Specific Comments
R probably has more bugs than Splus, but these are usually fixed a lot
faster than bugs in Splus. New versions come out about once every 4-6
months, so you need to be prepared to update/re-install often. R *is* free,
so just try it and see if it is adequate.
R is faster
R has a smaller footprint.
S-Specific Comments
The problem with Splus is that it is moving more and more towards a menu
system instead of a command language. So, for instance, I now find the help
in Splus useless because it gives me information on which menus to click on
and I cannot easily find the explanations for the command language. I find
in general the command menu in R more fun to use.
General
I have been using S-Plus for 2 years, and I recently switched to R. I really
like it, I think I like it even more than S-Plus, everything is free, well
maintained and easy to use.
I switched to R and have found it better than S+ in a number of ways,
irrespective of the fact that you can't beat the price. I've also think
that the email help list is more informative as well as more fun.
Depends a lot on the type of work you do. Some things are better in R, some
not as good, but mostly very similar.
In fairness to Insightful, their maintenance fee is quite modest. I still
use Ver 3.4 for what works best in Splus and R for what works best there
(and occasionally Genstat).
Is it time to switch to R? Yes (IMHO).
I tried to motivate Insightful to counter my University's SPSS infestation
with a competitive site licensing scheme, but they were/are apparently
uninterested. Consequently, I was unable to afford a single user license
and stopped using SPLUS after 10 years.
While my subsequent migration to R was effortless and quite pleasant, I
regret that some of my less practiced colleagues would-not or could-not
forswear the vending machine approach to executing statistical tests enabled
by menued programs like SPSS and SPLUS.
I'd strongly recommend trying it out. After all, it is free and it's not
hard to install and maintain. I couldn't advise you to abandon S-Plus,
because it's possible that you will need it for certain things. Perhaps you
could keep S-Plus for a while and do side-by-side comparisons, then you can
decide if you can let go of S-Plus.
|