Poisson perturbations (Q4268397)
From MaRDI portal
!
This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:
scientific article; zbMATH DE number 1354088
| Language | Label | Description | Also known as |
|---|---|---|---|
| default for all languages | No label defined |
||
| English | Poisson perturbations |
scientific article; zbMATH DE number 1354088 |
Statements
Poisson perturbations (English)
0 references
31 October 1999
0 references
Stein's method
0 references
signed compound Poisson measure
0 references
total variation
0 references
coupling
0 references
sums of random variables
0 references
compound measures
0 references
0 references
0 references
0.8889767
0 references
0.8801922
0 references
Signed compound Poisson measures have been used to make close approximations to the distributions of independent integer valued random variables. In this paper, measures having a generating function of the form \(\exp \{\mu(z-1)+ 2^{-1}a (z^2-1)\}\), \(\mu>0\), \(a\in {\mathfrak R}\), are considered and Stein's method is used to prove approximations in total variation. For negative values of \(\alpha\) the measure is a signed measure. For positive values of \(\alpha\) it is the convolution of a Poisson distribution with the distribution of \(2Z\) where \(Z\) is itself another Poisson distribution. The authors show that measures in this class are often close to other known distributions (like the binomial and the negative binomial distributions). Then the measures are used to approximate in total variation the distribution of the sum of independent integer valued random variables. The approximations are better than those existing in the literature for sums of Bernoulli variates while the approximations for other distributions are new. For the case when \(0<a<1\) the family of measures induced comes as near as possible to a family of translates of the Poisson distribution by any real displacement, while for large values of \(\mu\) (the parameter of the Poisson distribution in the convolution representation), the family is very close to a negative binomial distribution suitably defined. Lastly approximations for sums of dependent integer values variables are provided.
0 references