Quasi Most Powerful Invariant Goodness‐of‐fit Tests (Q4828208)
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 2118814
| Language | Label | Description | Also known as |
|---|---|---|---|
| default for all languages | No label defined |
||
| English | Quasi Most Powerful Invariant Goodness‐of‐fit Tests |
scientific article; zbMATH DE number 2118814 |
Statements
Quasi Most Powerful Invariant Goodness‐of‐fit Tests (English)
0 references
24 November 2004
0 references
goodness-of-fit test
0 references
Laplace method
0 references
location-scale families
0 references
maximal invariant
0 references
most powerful invariant test
0 references
0 references
0.9148229956626892
0 references
0.8310169577598572
0 references
0.7896091938018799
0 references
0.7845015525817871
0 references
The paper deals with the problem how to determine the true family of densities from which the sample \(X=(x_1,\dots,x_n)\) of i.i.d. observations was collected in the case of two location-scale families of densities; i.e., it focuses on the problem of testing \(H_0:f(x)=(1/\sigma)f_0((x-M)/\sigma))\) versus \(H_1:f(x)=(1/\sigma)f_1((x-M)/\sigma))\), where \((M,\sigma)\) are unknown location-scale parameters and \(f_0(\cdot), f_1(\cdot)\) are two given densities. The authors develop an approximation for the most powerful invariant (MPI) test; their approximation is based on the Laplace method for integrals and yields a very accurate approximation of the density of a maximal invariant. A new test, called QMPI (quasi-MPI), is derived. A method of computation of approximate \(p\)-values is presented. The power of QMPI tests for finite \(n\) is investigated and in many situations it occurs to be very close to (and may coincide with) that of the MPI test. The asymptotic behaviour of the QMPI test statistics under \(H_0\) is investigated. Many examples are worked out and they demonstrate that the new test can be applied to a much wider set of pairs of densities than previously was possible. By using versions of the Laplace method, the approach is extended to goodness-of-fit tests for residuals in regression and some multivariate distributions. A small simulation study confirms the theoretical results.
0 references