Random walks with similar transition probabilities (Q1874180): Difference between revisions

The author considers random walks on the nonnegative integers. It is assumed that the one-step transition probabilities \(p_{j,j+1}\) and \(p_{j+1,j}\) are positive and that \(p_{j,j}\) are nonnegative, for \(j\geq 0\), and that \(p_{ij}=0\) for \(|i-j|>1\). It is further assumed that \(q_0:=1-p_{0,1}-p_{0,0}\geq 0\); if \(q_0>0\), then the walk has an (ignored) absorbing state, labelled \(-1\). A random walk \(\tilde{\mathcal X}\) is called \(\alpha\)-similar to a random walk \({\mathcal X}\) if there exist constants \(C_{ij}\) such that the corresponding \(n\)-step transition probabilities satisfy \(\tilde{P}_{ij}^{(n)}=\alpha^{-n}C_{ij} P_{ij}^{(n)}\) for \(i,j\geq 0\) and \(n\geq 1\). Necessary and sufficient conditions are given for the \(\alpha\)-similarity of two random walks in terms of the one-step transition probabilities and also in terms of the corresponding spectral measures of Karlin and McGregor. The similarity concept has been previously considered for Markov processes; see e.g. \textit{P. K. Pollett} [J. Appl. Probab. 38A, Spec. Vol., 53-65 (2001; Zbl 1012.60067)].