scientific article; zbMATH DE number 4060467

From MaRDI portal
Publication:3795013





Let \(E=\{1,2,...,n\}\) and \(X_ t(t=0,1,...)\) be a homogeneous Markov chain on E with transition probability matrix \((P_{ij})\). Divide E into disjoint subsets \(A_{\alpha}\), \(\alpha =1,2,...,m\) and define a new process \(\hat X_ t\) as follows: \(\hat X_ t=A_{\alpha}\) if \(X_ t\in A_{\alpha}\), \(\alpha =1,2,...,m.\) Let R be a set of probability measures on E. We say that \(X_ t\) is an aggregation according to the partition \(\{A_{\alpha}\}\) and the set R if there exist an \(m\times m\) matrix \(\hat P\) and a vector \(\hat p(t)\) such that for all initial distributions \(p(0)\in R\), \[ \hat p_{\alpha}(t)= {\mathbb{P}}\{\hat X_ t=A_{\alpha}\}\;and\;\hat p(t+1)=\hat p(t)\hat P,\quad t=0,1,2,... \] If in addition \(\hat X_ t\) is Markovian, then \(X_ t\) is called a lumping with respect to \(\{A_{\alpha}\}\) and R. Some characterizations of \(X_ t\) to be a lumping are presented.




Cited in
(37)








This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3795013)