On goodness-of-fit tests for the Bell distribution (Q2175219): Difference between revisions
From MaRDI portal
Changed an Item |
Changed an Item |
||
Property / describes a project that uses | |||
Property / describes a project that uses: DLMF / rank | |||
Normal rank |
Revision as of 13:15, 28 February 2024
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | On goodness-of-fit tests for the Bell distribution |
scientific article |
Statements
On goodness-of-fit tests for the Bell distribution (English)
0 references
28 April 2020
0 references
A random variable \(X\) is said to have a Bell distribution with parameter \(\theta>0\), written \(X\sim\mbox{Bell}(\theta)\), if \[ P(X=k)=\frac{\theta^k e^{-e^\theta+1}B_k}{k!},\qquad k=0,1,2,\ldots, \] where \(B_k\) are the Bell numbers. The authors propose a goodness-of-fit test for the null hypothesis \(\mbox{H}_0\mbox{: }X\sim\mbox{Bell}(\theta)\mbox{ for some }\theta>0\), against the alternative \(\mbox{H}_1\mbox{: }X\not\sim\mbox{Bell}(\theta)\mbox{ for any }\theta>0\). The test statistic used is based on the observation that the PGF of the Bell distribution is the unique solution of the differential equation \(g^\prime(t)-\theta e^{\theta t}g(t)=0\) for all \(t\in[0,1]\). The asymptotic distribution of the test statistic is derived when the data have a Bell distribution; since this asymptotic distribution depends on the unknown parameter \(\theta\), a bootstrap procedure is used to assess statistically significant values of the test statistic. A simulation study and applications to data illustrate the performance of the proposed test compared to other goodness-of-fit tests available in the literature.
0 references
count data
0 references
goodness-of-fit tests
0 references
parametric bootstrap
0 references
empirical probability generating function
0 references
Bell distribution
0 references