Compound random mappings (Q4804728)

The authors introduce the compound random mapping model as follows. Let \(W_1,W_2,\dots\) be a sequence of independent, identically distributed and nonnegative random variables. For an integer \(K\geq 1\), let \(N= N(K):= \sum^K_{i=1} W_i\) and let \(\pi\) denote the random probability measure on \(\{1,2,\dots, K\}\) given by \(\pi= (1/N)(W_1, W_2,\dots, W_K)\) if \(N\neq 0\) and \(\pi= (1/K,1/K,\dots, 1/K\}\) otherwise. Further, denote by \({\mathcal M}_K\) the set of all mappings from \(\{1,2,\dots, K\}\) into itself and define the compound random mapping \(T_K(\pi)\) on \({\mathcal M}_K\) by \[ P\{T_K(\pi)= f\}= \int_{(R^+)^K} P\{T_K(\pi)= f\mid(W_1,\dots, W_K)= (w_1,\dots, w_K)\} dF(w_1,\dots, w_K) \] for any \(f\in{\mathcal M}_K\), where \(F\) is the joint distribution function of \((W_1,\dots, W_K)\). Let \(G_K(\pi)\) denote the random digraph associated with \(T_K(\pi)\). The variables \(W_1,W_2,\dots\) can be viewed in this model as relative ``weights'' on the vertices \(\{1,\dots, K\}\) of \(G_K(\pi)\). If \(W_i\equiv \text{const}\geq 0\), one obtains the classical uniform model of random mappings with digraphs denoted by \(G_K\). The authors assume that \(W_1,W_2,\dots\) have Poisson distribution with common mean \(\lambda> 0\) and call the corresponding model \(T_K(\lambda)\) with a digraph \(G_K(\lambda)\) the Poisson compound mapping. They show that, for every \(\lambda>0\), the joint distribution of the normalized order statistics of component sizes in \(G_K(\lambda)\) has the same limiting distribution as the joint distribution of the normalized order statistics of the component sizes in \(G_K\), i.e. the Poisson-Dirichlet \((1/2)\) distribution. The limiting distribution for the size of a component in \(G_K(\lambda)\) containing a given vertex is also found.