A martingale approach for Pólya urn processes (Q782829)

This paper studies a martingale approach for Pólya urn processes started with \(\alpha\) red balls and \(\beta\) white balls. The replacement matrix has the same row sum \(S\). Let \(\tau=\alpha+\beta\ge1\) and \(\tau_n=\tau+nS\) be the number of balls inside the urn at time step \(n\). The number of red balls at time step \(n\) for the Pólya urn process is denoted by \(X_n\) and let \(M_n=X_n/\tau_n\). It is shown that the process \(M_n\) converges to a random variable \(M_{\infty}\) almost surely and in any \(L^p\) space for \(p\ge1\). Moreover, the limit variable is shown to follow a beta distribution with parameters \(\alpha/S\) and \(\beta/S\).