scientific article; zbMATH DE number 3967562

From MaRDI portal
Publication:3733879





Suppose that we have a simple stationary homogeneous Markov chain with s states \(E_ i\) \((i=1,...,s)\), transition probability matrix \(\| p_{ij}\|_{i,j=1,...,s}\), and vector of initial probabilities \((p_ i)_{i=1,...,s}\). According to \textit{A. Ya. Khinchin} [Usp. Mat. Nauk 8, No.3(55), 3-20 (1953; Zbl 0050.354); English translation in Transl., II. Ser., Am. Math. Soc. 12, 181-197 (1959)], the entropy of such a chain is defined by the formula \[ H=\sum^{s}_{i=1}p_ i\ln p_ i-\sum^{s}_{i,j=1}p(ij)\ln p(ij), \] where \(p(ij)=p_ ip_{ij}\) is the probability that the chain \((E_ i,E_ j)\) occurs. Further, let \(\hat p_ i\) be the frequency with which the state \(E_ i\) is hit and \(\hat p(ij)\) the frequency with which the chain \((E_ i,E_ j)\) occurs in N trials. It was established by \textit{E. V. Misevichyus} [Litov. Mat. Sb. 6, 394-495 (1966; Zbl 0142.159)] that the statistical estimator \[ \hat H=\sum^{s}_{i=1}\hat p_ i\ln \hat p_ i-\sum^{s}_{i,j=1}\hat p(ij)\ln \hat p(ij) \] of H constructed from a sample of size N is asymptotically normal as \(N\to \infty\). We establish the class of limit distributions for an estimator of the entropy of the Markov chain indicated above with two states when the \(p_{ij}\) (and hence also the \(p_ i)\) depend on N.











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 Q3733879)