[Scipy-tickets] [SciPy] #486: Built-in Poisson CDF and PMF function slower than using built-in gamma

SciPy scipy-tickets@scipy....
Fri Aug 24 12:54:24 CDT 2007


#486: Built-in Poisson CDF and PMF function slower than using built-in gamma
-------------------------+--------------------------------------------------
 Reporter:  danhs        |       Owner:  somebody                                                       
     Type:  defect       |      Status:  new                                                            
 Priority:  normal       |   Milestone:                                                                 
Component:  scipy.stats  |     Version:  0.5.2                                                          
 Severity:  normal       |    Keywords:  probability distribution, scipy.stats, speed, poisson, binomial
-------------------------+--------------------------------------------------
 I'm not sure how or why, but scipy.stats.poisson.cdf is much slower than
 using using the incomplete gamma function as defined on wikipedia
 http://en.wikipedia.org/wiki/Poisson_distribution.

 That is poisson_cdf(k, lam) = scipy.special.gammaincc(floor(k+1), lam).

 The same is true for the binomial cdf:
 binom_cdf(s, N, p) = scipy.special.betainc(N-floor(s), 1 + floor(s), 1-p).

 And the binomial pmf:
 binomPmf = lambda s, N, p: scipy.misc.comb(N, s) * p**s * (1-p)**(N-s)

 I thought the situation with the normal distribution was similar, but to
 be frank, I don't have that code written and I'm not sure I tested it.

 This was run on Windows XP machine using Python 2.5.1. Scipy version used
 is 0.5.2 Numpy version is 1.0.3

-- 
Ticket URL: <http://projects.scipy.org/scipy/scipy/ticket/486>
SciPy <http://www.scipy.org/>
SciPy is open-source software for mathematics, science, and engineering.


More information about the Scipy-tickets mailing list