| Ticket UUID: | 1861700 | |||
| Title: | Submission: Beta dist and Incomplete Beta Function | |||
| Type: | RFE | Version: | None | |
| Submitter: | erickb | Created on: | 2008-01-01 15:20:33 | |
| Subsystem: | math | Assigned To: | arjenmarkus | |
| Priority: | 5 Medium | Severity: | ||
| Status: | Closed | Last Modified: | 2013-07-04 17:25:25 | |
| Resolution: | Not Applicable Here | Closed By: | ||
| Closed on: | ||||
| Description: |
Hi, I've implemented the Beta distribution and also provided an implementation of the incomplete Beta function (which is the CDF for the Beta distribution). I've tested the incomplete Beta function quite a bit (as shown), and think that it is now robust. However, I probably didn't make the most efficient implementation. An alternative would be the collection of algorithms in DCDFLIB (http://people.scs.fsu.edu/~burkardt/f_src/dcdflib/dcdflib.html), but it is quite involved -- if this implementation is sufficient, so much the better. Eric | |||
| User Comments: |
arjenmarkus added on 2008-01-18 16:39:13:
Logged In: YES user_id=400048 Originator: NO Incorporated this. One thing to think about: pdf-beta a b 0.0 or 1.0 are edge cases that should give either 0 or infinity (depending on a and b). Right now, they give 0 only. erickb added on 2008-01-12 07:15:58: File Deleted - 260418: Logged In: YES user_id=816411 Originator: YES This is a better implementation for the beta distribution. First, I extended the useful range for the pdf. Second, I implemented a much faster algorithm for the incomplete beta function, using a continued fraction from Abramowitz & Stegun. (I used the GNU Scientific Library for inspiration, but ended up using a different continued fraction expansion than they used, so there's no GPL concern here. The speeds are comparable.) File Added: beta.zip erickb added on 2008-01-12 07:15:57: File Added - 261872: beta.zip erickb added on 2008-01-02 09:36:39: File Deleted - 260416: File Added - 260418: beta.zip Logged In: YES user_id=816411 Originator: YES File Added: beta.zip erickb added on 2008-01-02 08:50:02: File Deleted - 260397: File Added - 260416: beta.zip Logged In: YES user_id=816411 Originator: YES File Added: beta.zip erickb added on 2008-01-02 05:27:40: File Deleted - 260396: File Added - 260397: beta.zip Logged In: YES user_id=816411 Originator: YES File Added: beta.zip erickb added on 2008-01-02 05:15:05: File Deleted - 260371: erickb added on 2008-01-02 05:15:02: File Added - 260396: beta.zip Logged In: YES user_id=816411 Originator: YES Attaching a new version. Fixes an edge case that can make the routine hang. Also introduces a short-cut in case of integer value for b. File Added: beta.zip erickb added on 2008-01-02 01:22:12: File Deleted - 260370: erickb added on 2008-01-02 01:22:11: File Added - 260371: beta.zip Logged In: YES user_id=816411 Originator: YES File Added: beta.zip erickb added on 2008-01-02 01:13:43: File Deleted - 260368: File Added - 260370: beta.zip Logged In: YES user_id=816411 Originator: YES File Added: beta.zip erickb added on 2008-01-02 01:01:18: File Deleted - 260358: File Added - 260368: beta.zip Logged In: YES user_id=816411 Originator: YES I'm attaching a version of the incomplete beta function that is significantly faster (about 10 faster for some values) than the original version I posted. File Added: beta.zip erickb added on 2008-01-01 22:27:53: File Deleted - 260356: File Added - 260358: beta.zip Logged In: YES user_id=816411 Originator: YES File Added: beta.zip erickb added on 2008-01-01 22:20:33: File Added - 260356: beta.zip | |||
Attachments:
- beta.zip [download] added by erickb on 2008-01-12 07:15:57. [details]