Immortal branching Markov processes: Averaging properties and PCR applications. (Q1879856): Difference between revisions

The immortal branching Markov process is a discrete-time branching process with immortal particles characterized by their types and states. A particle \(x\) gets its type \(t(x)\) and state \(s(x)\in R\) at the birth moment \(\tau_x\) and never changes them. At each moment \(n\geq\tau_x\), the particle \(x\) produces a random number of children. The overall population \(S_n\) at time \(n\) performs a classical branching process, that is the offspring size of a particle has distribution which is independent of the state and type of the parent particle. However, the type and state of newborn particles depend on the type and state of their parent particle in a Markov way. Let \(\eta_n\) be the empirical law of \(S_n\) if the process starts by \(S_0\) particles of the same state and type and \(\eta^*_n\) be the limit of \(\eta_n\) as \(S_0\to\infty.\) Let \(M_n\) and \(M^*_n\) be the random variables with the laws \(\eta_n\) and \(\eta^*_n\), respectively. The author gives explicit estimates from above and below for the difference \(EM_n-EM^*_n\) in terms of the characteristics of the underlying branching process and shows that \(| \text{Var}(M_n)-\text{Var}(M_n^*)| \leq \Gamma/S_0\) where \(\Gamma\) is a finite constant independent of \(n\) and \(S_0\). These results (and some others of the same spirit) are used in the paper to study the variability of genetic sequences produced by polymerase chain reactions (PCR).