making a case for FORTRAN

===========================================================================

Re: making a case for fortran Conf: (0) comp.lang.

---------------------------------------------------------------------------

Quote:

>I usually never read the fortran group, being involved in the c and object

>oriented worlds instead, but i might be induced to change now.

>I am planning to attend graduate school in statistics and i am looking for

>a general purpose language that once i get good at will do everything i

>need.

>i need:

>easy and powerful matrix math.

>good numerical anlysis routinues

>good error trapping and debugging.

>good graphics, the kind i can whipp together and paste into a paper.

>the ability to be able to build off of what i've done before, ie build

>routinues in one class and use them for all others.

>does Fortran get me there? if so which compiler do you recommed. i'll go

>out a buy a new computer in the fall and i was thinking of a powerPC or a

>windows clone. pluses if its also portable to a unix box back and forth.

>thnaks for your advice.

>joe alotta.

My impression from colleagues engaged in statistical computing research, is

that FORTRAN continues to be the core number crunching language mostly because

of the vast existing code libraries of esoteric & arcane statistical &

mathematical routines. C is used for graphical interfaces, vizualization and

by some for numerical work. APL has enjoyed a minor renaissance in some

quarters for its superior matrix math.

For no other reason, than to decipher existing FORTRAN library routines, you

wouldn't go far wrong to learn basic FORTRAN [heck it's one of the easier

languages to learn], and it certainly would not have to cost much, i.e.,

there's at least one free F77 compiler and a FORTAN to C converter/compiler.

For it's advanced modern statistical routines and for it's superior graphics,

the S-PLUS stat package has become a premier research tool. It has the added

advantage of being WATCOMM compatible, i.e., you can load your own C or

FORTRAN routines; hence, get around your graphics hard copy problems.

In short, no single language or statistical package will guarantee what you

ask. As a serious statistical researcher you'll be exposed to a variety of

statistical software in a variety of languages. If you're numerically

inclined the more you know the better. At the other extreme, if you're

inclined to theoretical abstraction, your computing might be confined mostly

to composing elegant mathematical treatises in Tex [frankly there's a lot to

be said for being a pencil pusher / symbol thrasher whose main computing need

is preparing manuscripts, but good jobs are few and far between].

Good Luck.

Byron Bodo

Adj Prof, Stats & Actuarial Sci

The University of Western Ontario

--

Byron Bodo 240 Markham St. tel: (416) 967 7286

|> |> Toronto, ON fax: (416) 967 9004