Good Statistical Packages

[Marc Schwartz]

Matthew Saltzman wrote:
> On Tue, 22 Mar 2005, Alexander Volovics wrote:
> 
>> On Mon, Mar 21, 2005 at 05:11:55PM -1000, Ping-Wu Zhang wrote:
>>
>>> Could anyone recommend any good statistical package(s)?  Does anyone
>>> know whether there is any SPSS or SAS clone?  Thanks.
>>
>>
>> If the license fee is not an objection ($615.00 for single user
>> with Base Documentation Set) I highly recomend Stata:
>> http://www.stata.com/
>> Maybe you can get you company/institute or university dept. to order it
>> for you. It is a bit easier to master than R and still cheaper than SAS.

The OP did not indicate academic status, so just in case:

Stata 8 SE current commercial single user license fees are $1,345 U.S.
with full documentation and $1,215 U.S. with base documentation.

Those are direct from the web site.

> If the license fee is not an issue, I beleive that SAS is available for
> Linux as well (although Alexander may be correct about the relative cost).

A single processor "commercial" (as opposed to academic) use license for
Linux for SAS is $7,800 U.S. for BASE and $4,200 U.S. for STAT. Those
are the first year costs for the minimum two SAS components that would
be required.

Keep in mind that SAS is licensed on an annual basis. Subsequent annual
renewals would be $2,560 U.S. for BASE and $1,390 U.S. for STAT.

If you wanted SAS' GRAPH component, that would be $3,900 U.S. for the
first year and $1,340 U.S. for subsequent renewals.

A lot of people I know who use SAS for analysis use S-PLUS and/or R for
graphics, simply because SAS GRAPH is a PITA to use.

>> However the advice depends on what you need it for.
>> Any package can produce simple summaries, graphs and the classical
>> analyses. However if you need things like certain random effects models
>> you might only be able to find the necessary tools in SAS for
>> example. (Programming something yourself in R is not very enjoyable).

I agree that one needs more information for a particular recommendation
to be made here relative to functionality. For example, if one needs
extensive exact methods, then StatXact (or the PROC STATXACT add-on
module for SAS) would be the way to go. StatXact is not available on
Linux however.

Beyond the base R installation, which includes BTW, functionality for
mixed effects models, there are ~500 add-on packages available to cover
many expanded functions.

Programming in R is a breeze as compared to programming in SAS, having
been a SAS user for a number of years. If one needs to implement some
new algorithm, there are interfaces for R to FORTRAN and C if needed and
a large community available for help.

That all being said, one needs to determine functionality requirements
and budgets to aid in the decsion making process.

HTH,

Marc Schwartz