Re: Statistics/graphing programs for scientists?
> The best statistical pagacke is IMHO R.
>
> -Egon
>
> On Thu, 11 Mar 1999, William Park wrote:
>
> > On Thu, Mar 11, 1999 at 08:29:38AM -0600, rich wrote:
> > > Hello all,
> > >
> > > I'm just about to get my doctorate in neuroscience,
> > > and I have have several large databases essential for my dissertation.
> > > For statistical analysis, I use Statistica for windows, and for graphing
> > > my data, I use SigmaPlot for windows. A call to all scientists out there
> > > - are there any native X-based programs that are as good as these?
> > > Although these programs are excellent, I would rather not trust my
> > > dissertation to the OS I have come to call Sir Crash-a-lot... My only
> > > other option is to use a windows emulator (like WINE)...
> > >
I agree the best statistical package is R, but it is best in the same
way Debian is the best Linux.
R demands some time before you get much a lot from it.
It has been called the Maseratti while SAS has been called the Ford.
R includes a full programming language.
While I use the graphics of R, I am uncertain of how well it makes them
for paper copies.
You learn to use R use either a 60 page document on R,
or a book like "Modern Applied Statistics with S-PLUS".
A really neat application, xgobi, interfaces with R.
xgobi is a graphical tool.
With it you can spin three dimensional objects;
in fact you can graph and follow higher dimensional objects,
though its weird and as they say, "viewing higher dimensions has a usefulness
related to your brain's abilities."
It's weird seeing 4 axes on a 2-dimensional graph.
Rotation helps perception greatly in 3 dimensions, but I'm not ready for 4.
I have brushed some points to green in 2-dimensions,
then viewed spining 3-dimensions with the green then showing.
xgobi makes all this easy: click the type of graphs (3-D rotation,
2-D) you want to view,
click from the various variables shown for your data set.
>From within R, I start xgobi by merely entering
xgobi(my-data-frame)
R is big on data-frames, which often have a first line "header"
defining your variables when read through "read.table".
Debian has the xgobi package that you can use with R, or on its own,
though it wants a dataset before it starts.
xgobi started at AT&T, with one author now working at
Iowa State, and a contributor working at George Mason University.
R is a GPL software that does almost identically what SPlus does.
You can buy SPlus for Linux: $4995.00.
R has some nice features and SPlus has some nice user friendly features,
but most programs work on either software.
There is also the graphics package gnuplot that some people use;
Octave (the MatLab clone) uses the gnuplot package to implement its graphics.
Since you are getting your PhD, you must be at a University.
Doesn't your university have a computer you can log into.
The Universities I am familiar with have the big statistical packages:
SAS, SPlus, SPSS, and maybe BMDP.
And the universities I know often have statistical consulting, free
to faculty and students.
They may have recommendations for your university.
I occasionally log into a Sun computer 30 miles away,
then bring up an X-windows SAS session on my Debian Linux.
Oh, you said your datasets are large.
Using the options -n and -v in R I have accomodated datasets with
30,000 records and 20 variables on a 128MB computer.
I have yet to run out of memory, but R/SPLUS do their computing in memory
rather than on/off the disk drive.
That makes R very very quick (it is a Mazaratti) but limits itself away
from some very large datasets.
While R will plot in X-windows,
all your programming is text based: no graphical programming interface.
--
Jim Burt, NJ9L, Fairfax, Virginia, USA
jameson@mnsinc.com http://www.mnsinc.com/jameson
jameson@pressroom.com (703) 235-5213 ext. 132 (work)
"A poor man associating with a rich man will soon be too poor
to buy even a pair of breeches." --Chinese Proverb
Reply to: