Log in

Graphical Models of Probability
Bayesian Networks, Decision Networks, and Probabilistic Relational Models
Hi, I am new to this forum. I am a mathematical sciences major… 
28th-May-2005 12:29 pm
Hi, I am new to this forum. I am a mathematical sciences major planning on taking elective statistics for the major, and hoping to get a few relevant questions answered:

Is Java a good programming language for statisticians? Numerical analysts? I noticed that this community is predicated on it.

EXTRA CREDIT-- I had a discrete mathematics exam, and I was able to probably earn partial credit on everything except this question: You have six coin tosses, and it's given that three heads and three tails are resultant. How many possible orders are there?
28th-May-2005 07:40 pm (UTC)
Several of the PhD Stats programs I interviewed with liked the fact that I knew Java, but I think it's best just to be familiar with programming in general, whether it's C++ or Java.... or SAS, which really isn't a programming language but it's used to do statistical programming a lot.

6! / (3!3!) = 20
28th-May-2005 07:47 pm (UTC)
OK, now I feel stupid, I remember doing that permutation in my homework and then forgetting all of it. It goes without saying that that's no good for math majors.

You're a soon-to-be Ph. D. statistician? Since I am physics-, chemistry-, finance-, and computer-illiterate, I have always figured any mathematical training I can have to my credit will tend toward finance. My university has a Ph. D. program in mathematical and empirical finance in the stats department. I have a finance professor-mentor (part time), who tells me that that would be interesting but hard to get a job in. Without further preamble, are you planning on being a teacher, or are you concerned with and about finding a job in industry?
28th-May-2005 08:36 pm (UTC)
I haven't done a problem like that in a while, so I had to pause for a moment also.

I'm planning on eventually steering myself over into Biostatistics, where I'll probably do some type of clinical pharmaceutical research while working on my PhD...probably / hopefully in a Bayesian (as opposed to classical) context.

I probably will teach someday, hopefully at the university graduate level since I don't have much patience for people ages 12 to 22. But we'll see. I'll probably stay in industry for as long as my life allows.
28th-May-2005 07:55 pm (UTC)
You know, I have to say that chicks who dig math are the shit, but for some reason I'm not surprised to see "Volkswagon Jetta" in your interests lists. ;)
28th-May-2005 08:32 pm (UTC)
Ehn, I used to have one, but then I was rear-ended at 60 mph by some illegal immigrants. So now I don't have one. I have a Toyota Matrix, with which I also have a love-hate relationship, for a variety of reasons.

Jettas have become too popular anyway. And their reliability isn't so great... and I'm all about practicality, more than what is "cool" or anything like that.
28th-May-2005 10:20 pm (UTC)
I'm horrible with non sequiturs, but I figured it was more polite to comment here where there's not a whole lot of discussion than in banazir's LJ. Their popularity is why I pick on them. That and I am kind of snobbish and see them as being kind of generic. Plus, back when I owned a twin turbo RX-7, people in Jettas would try to race me and at that time I would make fun of people who would race cars that're nowhere near the same class as theirs.
28th-May-2005 10:38 pm (UTC)
I generally make fun of anyone who races anything, or anyone who drives anything other than a purely practical car.
29th-May-2005 05:13 am (UTC)
Of course you do.
29th-May-2005 05:13 am (UTC)
There are many more numerically suited languages out there. Python is particularly well-suited to your position, as it has one of the gentlest learning curves, numerical extensions such as numpy, intepreted portability (there's even a Java version of the interpreter!), and a host of extensions to facilitate for the collection and pre-processing of data from many sources on the 'net.
29th-May-2005 08:49 pm (UTC)
By "numerically suited," were you talking about numerical analysis, or even prob/stat, which is, duh, the purpose of the board?
11th-Jul-2005 04:09 pm (UTC)
Learning statistical packages is helpful like SAS, SPSS, Splus, minitab, etc.

The more you know, the better off you'll be.
3rd-Dec-2005 09:46 pm (UTC)
For statistics:

You need to learn the language that goes with your favourite stats package. Other than that, Perl is handy for reformatting data files. You might need a language to do some home-brewed code for, say, MCMC or simulations, and Perl would not be a great choice there. (It's interpreted.) I'd be tempted to use Mathematica, MATLAB, or R for those.

For numerical analysis:

Fortran is the traditional language and it's still the one to beat for serious numerical number-crunching. Fortran 77 is free (from the GNU project) but it's a rather old standard that does not support recursive data structures, recursive subroutines or pointers. Later versions overcome these limitations and suppport parallel processing, but these are not available free.

C and C++ are widely used, but I believe them to be seriously flawed. Pointers are used in an uncontrolled way that makes makes memory leak bugs very easy to create. Weak typing and manual memory management are also major sources of bugs.

Java has designed out the problems with C and C++ and is much superior. However, it is interpreted and so may not be fast enough for some major computational problems. I tested Java (used with a Just-In-Time compiler) versus C++ and found that C++ ran in half the time. I was very pleased with the performance of Java from this test. A performance penalty factor of 2 for the advantages of Java is a price that's well worth paying. For most purposes I would regard Java as entirely practical for mathematical computation.
3rd-Dec-2005 09:50 pm (UTC)
I appreciate it. My local college has MATLAB and Java courses for sure, and I am afraid I am going to have to refresh what Java skills I never picked up in the first place in university recently. It's all about numerical analysis, I don't need statistical computer skills yet, but I'd love to pick them up. Thanks for the reply.
9th-Dec-2005 12:35 pm (UTC) - MATLAB and Fortran
I should have said that MATLAB is great for numerical analysis. It was originally designed by Cleve Moler as a front end to a large library of numerical routines written in Fortran, but it took off as a product in its own right.

I looked up the GNU Fortran yesterday and found that they have moved it on quite a bit. The F77 compiler has been replaced by an F95 compiler. It seems it's not quite finished yet, but usable. So, F95 is now free, too!
This page was loaded Jul 24th 2017, 4:32 pm GMT.