When to use zero-truncated Poisson or zero-truncated negative binomial
87 views
Skip to first unread message
geyer
Oct 31, 2015, 5:54:31 AM10/31/15
to Aster Analysis User Group
The zero-truncated distributions are for when their variables cannot be zero given predecessor = 1. The usual pattern is
Ber 0-Poi
... -------> any foo -------> num foo -------> ...
where any foo == 1 if and only if num foo >= 1. The combination of the two arrows Ber followed by 0-Poi makes a zero inflated Poisson (and, analogously, the combination of Bernoulli followed by zero-truncated negative binomial makes a zero-inflated negative binomial).
The zero-truncated distributions are not for when there just happen to be no zeros for a count variable (because the mean is fairly large).
Ber 0-Poi
... -------> any foo -------> num foo -------> ...
where any foo == 1 if and only if num foo >= 1. The combination of the two arrows Ber followed by 0-Poi makes a zero inflated Poisson (and, analogously, the combination of Bernoulli followed by zero-truncated negative binomial makes a zero-inflated negative binomial).
The zero-truncated distributions are not for when there just happen to be no zeros for a count variable (because the mean is fairly large).
Reply all
Reply to author
Forward
0 new messages