[ot-users] cdf of mixture distribution

François Deheeger francois.deheeger at free.fr
Wed May 7 22:40:25 CEST 2014


Hello OT users,

i have a question about mixture and their cdf computation. Is there any way
to control the precision of the computation ?

I have some use case where for some reason the computation seems to be lost
(or at least not as efficient as expected).

Do you have any advice on how to control the following example ?

import openturns as ot
import numpy as np

for i in [1, 2, 5, 8, 10]:
    dist1 = ot.Weibull(45000., 2.5)
    dist2 = ot.Weibull(i * 12500., 2.5)
    print i
    diff = dist1.getSample(10000) - dist2.getSample(10000)
    print np.sum(np.array(diff) < 0) / 10000.
    print ot.RandomMixture([dist1, dist2], [1., -1.]).computeCDF(0.)

The first estimation if the probability is just to get an idea of the
result. I thought at the beginning that i was looking for too small
probabilities but here it is not the case.
(i put 10 in the list, but i didn't even get the result for i=8 !)


Thanks,

-- 

François Deheeger
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://openturns.org/pipermail/users/attachments/20140507/02350b4f/attachment.html>


More information about the users mailing list