[Octave-bug-tracker] [bug #39309] [octave forge] (statistics) poisscdf fails for large arguments due to the algorithm in D9LGIT

[Nicholas Jankowski]

Update of bug #39309 (project octave):

                  Status:               Confirmed => Fixed                  
             Open/Closed:                    Open => Closed                 

    _______________________________________________________

Follow-up Comment #5:

looking at the current Statistics 1.4.3 with Octave 7.1.0- 

poisscdf is largely just a call to gammainc, which is in core octave, and has
undergone at least a few revisions in the last decade.

the comment #0, comment #1, comment #2 codes no longer produce fortran
errors:


comment #0:

>> poisscdf (604,604)
ans = 0.510819531737768
>> poisscdf (605,605)
ans = 0.510810590057861
>> poisscdf (4241,4241)
ans = 0.504083866960345
>> poisscdf (4242,4242)
ans = 0.504083385599947


comment #1:

>> poisscdf (100000,100000)
ans = 0.500841043099343
>> poisscdf (100001,100001)
ans = 0.500841038894170
>> poisscdf (1e10,1e10+1e5)
ans = 0.158657673645873


comment #2

>> poisscdf (10,10)
ans = 0.583039750192986
>> poisscdf (100,100)
ans = 0.526562198529999
>> poisscdf (1000,1000)
ans = 0.508409367168506
>> poisscdf (100000,100000)
ans = 0.500841043099343
>> poisscdf (100000000,100000000)
ans = 0.500026596152138
>> poisscdf (100000000000,100000000000)
ans = 0.500000841049293
>> poisscdf (100000000000000,100000000000000)
ans = 0.500000026770566
>> poisscdf (100000000000000000,100000000000000000)
ans = 84167.84956743612


soooo....  it seems that the original issue has largely been fixed. 

the last test did go way too fast, so I assume it did hit some internal
convergence error, although there were no fortran errors/warnings.  that part
is a bit concerning, as I would have hoped some non-convergence warning might
have been issued.  but again, that would seem to be an issue with gammainc and
not poisscdf. so I'm going to close this old bug report.



    _______________________________________________________

Reply to this item at:

  <https://savannah.gnu.org/bugs/?39309>

_______________________________________________
Message sent via Savannah
https://savannah.gnu.org/