Central and functional central limit theorems for a class of urn models (Q867092)

By a tail-based approach the authors establish certain central limit and functional central limit theorems for a class of two color urn models. More specifically, they consider an urn model in which one starts with \(s\) black/white balls, initially. At each step one ball is drawn at random and along with it \(\alpha\) \((> 0)\) balls of the same color and \(\beta\) \((> 0)\) balls of the opposite color are added. Let \(\delta=\alpha- \beta\) and \(\rho= \delta/(\alpha+\beta)\). Let \(W_n\) and \(B_n\) denote the number of white and black balls, respectively, just before the \((n+ 1)\)st draw. Among others, the following weak convergence theorems are shown: (i) in the case \(\rho< 1/2\), \[ t^{1/2-\rho} {W_{[nt]}- B_{[nt]}\over \sqrt{nt}}\Rightarrow {1\over\sqrt{1- \rho}}, \] (ii) in the case \(\rho= 1/2\), \[ \sqrt{t}{W_{[nt]}- B_{[nt]}\over \sqrt{[n^t]\log[n^t]}}\Rightarrow\delta W(t), \] (iii) in the case \(\rho> 1/2\), \({Y_n- B_n\over \prod^{n-1}_0 a_\nu(1)}@>\text{a.s.}>>Y\) \quad and \quad \(n^{\rho-1/2}(Y_{[nt^{1/(2\rho-1)}]}- Y)\Rightarrow W(1/t)\) for some deterministic sequence \(\prod^{n-1}_0 a_\nu(1)\) depending on \(s\), \(\alpha\) and \(\beta\). Here, \(W(\cdot)\) denotes a Brownian motion.