| Ticket UUID: | 1838703 | |||
| Title: | cdf-normal gives div-by-zero for arg = 0 | |||
| Type: | Bug | Version: | None | |
| Submitter: | erickb | Created on: | 2007-11-26 13:42:57 | |
| Subsystem: | math :: statistics | Assigned To: | arjenmarkus | |
| Priority: | 5 Medium | Severity: | ||
| Status: | Closed | Last Modified: | 2009-11-16 15:20:56 | |
| Resolution: | Fixed | Closed By: | arjenmarkus | |
| Closed on: | 2009-11-16 08:20:56 | |||
| Description: |
Hello, The cdf-normal function gives a divide-by-zero error for the following call (which should return 0.5): ::math::statistics::cdf-normal 0 1 0 The problem appears to arise in Cdf-toms322, where the internal variables w and p become zero, and then one is divided by the other. Eric | |||
| User Comments: |
arjenmarkus added on 2009-11-16 15:20:56:
allow_comments - 1 Closing this issue - the new method avoids this erickb added on 2007-11-26 21:54:37: Logged In: YES user_id=816411 Originator: YES I can do the quick-and-dirty workaround no problem. Thanks! I just wanted to bring it to your attention. Take care, Eric arjenmarkus added on 2007-11-26 21:29:18: Logged In: YES user_id=400048 Originator: NO No, I assumed it was because that is a common mistake with converting Fortran or C code to Tcl. However, we are in some limit situation for the algorithm. The quick-and-dirty solution is treat x=0 as a special case. erickb added on 2007-11-26 21:21:36: Logged In: YES user_id=816411 Originator: YES Apparently not an integer problem: ::math::statistics::cdf-normal 0.0 1.0 0.0 gives the same error. arjenmarkus added on 2007-11-26 21:01:01: Logged In: YES user_id=400048 Originator: NO I will have a look at it - perhaps a problem with integer values (rather than floats)? | |||
Home
Blog
Docs
Downloads
Tickets
Timeline
Branches
Tags
Wiki