Inverse binomial



Post: #9

1) On both my emulator and my physical unit, inverse binomial with p=.5 stored in j, n=10 stored in k for .005 yields 10. I was expecting 0 or possibly 1.

.006 correctly produces an answer of 1.

2) On my emulator when trying large n such as 1 E 06, inverse binomial p=.5, n (k)=1 E 06 for .75 produces 500337. My physical unit produces "reset". The physical unit will come up with a proper result for .75=501288.

2) might be due to older version (3.1 3278)


Post: #10

Physical unit produces correct answer of 501288 for .995 (not the .75 that I accidentally reported).

Post: #11

I'm not the one to answer your question about the proper results, but "reset" is definitely not the intended answer. It looks like another internal stack overflow issue or the like.

Post: #12

Using the R statistical package I get a value of 1 for your first example

qbinom(.005,10,.5)

1

and 500337 for your second example

qbinom(.75,1e6,.5)

500337

Nick


Edited: 10 July 2013, 11:28 a.m.

Post: #13

Hi Richard,

Thanks for reporting. Your observation 1) is definitively a bug. Even 0.0051 Binom^(-1) returns 1. I guess the cause is some internal rounding or truncating.

d:-)


Post: #14

I think we should fix the memory/stack overflow/corruption bug 2 before jumping to any conclusions about the cause of 1.

- Pauli


Post: #15

I concur. Go ahead!

d:-)

Post: #16

The "Reset" proved to be the watch dog kicking in. I've added another call to service the watch dog in the code and the "Reset" seems to be gone. Maybe Pauli can find a way to speed up the calculation a little. It just took too long...


Forum Jump: