Question on libraries for scientific computing (statistics)

[TiagoPereira]

Dear all,

Although I am very new to the C programming language, I am really, but really happy with its capacities compared to standard statistical packages (much, much faster for large scale analyses!).

In this respect, I have searched around potential libraries for scientific computing, and found out: the GNU Scientific Library (GSL) as well as the Apophenia library.

Both provide an excellent number of statistical functions. However, there are several important functions in statistics that they still miss, and it is hard for a beginner like to implement them within a short period of time.

Hence, are you aware of any other potential libraries for scientific computing that might contain additional functions for statistical analysis? My concern is that, as a beginner in C programming, I may be missing some important resources.

Grateful for any tips.

Cheers!

Tiago

[grumpy]

You haven't been too specific .... look here.

If you're willing to look at commercially available libraries (you have to pay) look up NAG and IMSL libraries.

[TiagoPereira]

Thank you so much! This is a step further: I did not know that there were commercial libraries with statistical functions. My R world is too free to think about that.

The commercial prices are, however, compatible with academic purposes. Even though google always helps a lot, are there any tutorials/books on additional potential non-free libraries (for dummies)?

thanks again!

Tiago

[nimitzhunter]

Don't think you can get NAG for individual license.

[TiagoPereira]

Why's that?

[johnggold]

I have a very old book "100 Solved problems in statistics published by McGraw Hill".

Unfortunately the solutions are written in Basic, but it's not difficult to convert - mostly it's line for line conversion. I'm sure that there's a newer version out there.

What type of calculation are you looking for - statistics cover a wide range - from basic mean, standard deviation through linear and multiple regression, chi suare tests and so on.

If I've know the solutions you need, I can probably dig them out for you.

[nimitzhunter]

Why's that?

I wanted a student licence from NAG like a year ago. So I contacted this guy I know, he worked with the NAG people. And the NAG people told that they don't offer single student licence. I'm not sure about non-student licence, but I figure the price would be very steep.

[pheininger]

Don't think you can get NAG for individual license.

Why's that?

I wanted a student licence from NAG like a year ago. So I contacted this guy I know, he worked with the NAG people. And the NAG people told that they don't offer single student licence. I'm not sure about non-student licence, but I figure the price would be very steep.

Apparently NAG is the Numerical Algorithms Group.

[TiagoPereira]

Many thanks, guys!

John, I basically need everything is statistics. right now I am eager to find a function that computes the natural logarithm of large factorials. For example, in R it is impossible to compute factorials for numbers larger than 177. However, the natural logarithm of the factorial is very straightforward:


lnfactorial(418500)
4998752.3

If you have any function to compute that, please, let me know asap.

Cheers!

Tiago

[johnggold]

I haven't run that far.

Don't understand the question.

Which statistical problem are you trying to solve.

That would help me guide you.

[pianorain]

Shouldn't the following piece of code work:

double lnfactorial(int fact)
{
   double x = 0.0;
   for(int i = 1; i <= fact; ++i)
   {
      x += log((double)i);
   }
   return x;
}

That function returned 4998752.25384429 for lnfactorial(418500). I wouldn't be surprised by some floating-point inaccuracy, but given the magnitude of the numbers, it should relatively small.

[TiagoPereira]

Yes, sure, please, see my post:

How to compute the natural logarithm of a factorial: large numbers

[TiagoPereira]

Just to follow up on my previous post: the NAG library costs around U$2,500 (Brazilian price; we have some additional taxes) plus "maintenance" (U$400,00), whereas the IMSL is cheaper: U$ 195 (no additional taxes). Although the NAG C library can be less expensive depending on the country, I believe that its price is always above 2k.