The journal Applied Statistics used to have a section on algorithms, which are available online. Griffiths and Hill also published a selection of these algorithms in a book in 1985. But I can't find any modern equivalent of these resources anywhere. By modern, I mean something that has the updated versions of the algorithms, includes more recent algorithms, and hopefully, the implementation is in a more modern programming language. Does it exist?
$\begingroup$
$\endgroup$
2
-
$\begingroup$ Is there a particular algorithm you inquire about? $\endgroup$– usεr11852Commented Aug 21, 2014 at 7:41
-
$\begingroup$ Originally, I was looking for an online algorithm to calculate a moving variance. $\endgroup$– user765195Commented Aug 21, 2014 at 12:47
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
1
The field of Computational Statistics has moved a long way since 1985 and probably a single book would be far too restrictive for an overview like this. In some instances it has even changed names... Check for example the following paper presenting the top 10 data mining algorithms identified by the 2006 IEEE International Conference on Data Mining (ICDM) found here. While good, it is already slightly outdated as it misses the impact of Deep Neural Networks and Gaussian Processes had in the last years. As for implementations most researchers give out some implementations in their respective websites.
EDIT (based on OP's comments)
I am not aware of a specific book on these topics. The closest matches I can think of are the books: Computational Statistics by J.E.Gentle and the Springer Handbooks of Computational Statistics. Especially the first one deals with issue of algorithmic efficiency and a specific mention on one- and two-pass sample variance computations is done but I think it is much more mathematically inclined in comparison with the 1985 book you mention.
For more "programmatically inclined" resource I would point you directly to journals such as the ACM Transactions on Mathematical Software, the SIAM Journal on Scientific Computing and the Journal of Computational and Graphical Statistics. The Journal of Statistical Software might also prove helpful if you are looking for implementations (and it's free).
Finally regarding the specific moving variance calculation you have mentioned I have seen two moderately fresh (~2008) resources on the matter by Pebay and Terriberry respectively (who actually cross-reference each others work) but other than not much.
-
$\begingroup$ Thanks. I'm aware of the progress that has been made in data mining and machine algorithms. For those, there is in fact an explosion of resources. What I meant was computational issues in statistics. Something as simple as calculating variance is a nontrivial problem (see for instance Welford's algorithm.) $\endgroup$ Commented Aug 21, 2014 at 12:51