Central limit theorems for nonlinear hierarchical sequences of random variables (Q5949331)

Let a random variable \(x_0\) and a function \(f: [a,b]^k\to[a,b]\) be given. A hierarchical sequence \(\{x_n: n=0,1,2,\ldots\}\) of random variables is defined inductively by the relation \(x_n=n=f(x_{n-1,1},x_{n-1,2},\ldots,x_{n-1,k})\), where \(x_{n-1,i: i=1,2,\ldots,k}\) is a family of independent random variables with the same distribution as \(x_{n-1}\). A function \(f\) is called averaging if: (1) For all \(u_i\in [a,b]\) \((i=1,2,\ldots,k)\): \(\min_i u_i \leq f(u_1,u_2,\ldots,u_k)\leq \max_i u_i\); (2) \(f\) is monotone increasing; (3) for all \(u<v\) in \([a,b]\) and for any two distinct indices \(i_1,i_2\in \{1,2,\ldots,k\}\) there exist \(u_i\in\{u,v\}\) \((i=1,2,\ldots,k)\) such that \(u_{i_1}=u\), \(u_{i_2}=v\) and \(u<f(u_1,u_2,\ldots,u_k)<v\). The central limit theorem for a hierarchical sequence of random variables is proved if the set \(\{u: \text{ for any } \varepsilon, {\mathbf P}[|x_0-u|<\varepsilon]>0\}\) is contained in \([a,b]\) and consists of more than one point, and the averaging satisfies some smoothness condition. The classical central limit theorem falls into this scheme. The sequence of conductivities arising in a random resistor network on the so-called diamond hierachical lattice is also a hierarchical sequence (Kirchhoff law). As a corollary the central limit theorem is proved for a suitably normalized sequence of conductivities of a random resistor network of this kind.