Changepoint Analysis as a Method for Isotonic Inference
Let data be from a one-way layout \(y_{ij}=\mu_{i}+\epsilon_{ij},\;i=1,\ldots,K,\;j=1,\ldots, n_{i}\), where \(\epsilon_{ij}\) are independently distributed as \(N(0,\sigma^2)\). The authors introduce the concavity hypothesis \[ (\mu_2-\mu_1)(x_2-x_1)^{-1}\geq (\mu_3-\mu_2)(x_3-x_2)^{-1}\geq\dots\geq (\mu_{K}-\mu_{K-1})(x_{K}-x_{K-1})^{-1} \] with at last one strict inequality for \(x_1\leq\ldots\leq x_{K}\) and derive a slope change model as its corner model. Then the likelihood ratio test statistic for the slope change model is developed as an extension of the maximal contrast type test (max \(t\)). Similarly, the sigmoidicity hypothesis \[ (x_3-x_1)^{-1}\left((\mu_3-\mu_2)(x_3-x_2)^{-1}- (\mu_2-\mu_1)(x_2-x_1)^{-1}\right)\geq \] \[ \geq (x_{K}-x_{K-2})^{-1} \left((\mu_{K}-\mu_{K-1})(x_{K}-x_{K-1})^{-1}- (\mu_{K-1}-\mu_{K-2})(x_{K-1}-x_{K-2})^{-1}\right) \] and the corresponding inflection point model are considered, as well as the extended max \(t\) test derived as a likelihood ratio test for the latter model. The distribution theory of the extended max \(t\) test and recurrence formulas for the level probability and also for the power calculation are presented. Some considerations on the power of the test and an application to binomial data are given.
- Isotonic regression: Another look at the changepoint problem
- Bayesian isotonic changepoint analysis
- Inference in a model with at most one slope-change point
- scientific article; zbMATH DE number 4143268
- A unifying approach to the shape and change-point hypotheses in the discrete univariate exponential family
- A unifying approach to the shape and change-point hypotheses in the discrete univariate exponential family
- Multiplicity adjustment for temporal and spatial scan statistics using Markov property
- A hybrid test for the isotonic change-point problem
- Isotonic regression: Another look at the changepoint problem
- On the isotonic change-point problem
- Profile Analysis of 24‐Hours Measurements of Blood Pressure
This page was built for publication: Changepoint Analysis as a Method for Isotonic Inference
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4416167)
