Limit behavior of a compound Poisson process with switching between multiple values (Q2246205)

Let \(\delta =e^{i2\pi /3}\) and assume that \((Y_{k}(1),Y_{k}(\delta ),Y_{k}(\delta ^{2})),k\geq 1\) are i.i.d. random vectors. Let \(X_{i},i\geq 1\) denote i.i.d. random variables with \(P(X_{1}=\delta )=p\) and \(P(X_{1}=1)=1-p\). Let \(N(t),t\geq 0\) denote a Poisson process with intensity \(\lambda \). We assume that all variables are independent from each other. Let \(X_{0}\) denote a nonrandom quantity taking one of the values \(1,\delta \) or \(\delta ^{2}\). In this paper, the authors study the compound Poisson process with switching defined as follows: \(B_{b}(t)=\sum_{k=1}^{N(t)}Y_{k}(\Pi _{j=0}^{k-1}X_{j}),t\geq 0\). Setting \(T_{p}(t)=\Pi _{j=0}^{N(t)}X_{j}\), the bivariate process \( Q_{p}(t)=(T_{p}(t),B_{p}(t)),t>0\) and \(Q_{p}(0)=(X_{0},0)\) is a homogeneoous Markov process (Theorem 2.1). In section 3 the authors assume that \( EY_{1}(1)=EY_{1}(\delta )=EY_{1}(\delta ^{2})=m\) and consider the following normalized process \(L_{a}(t)=(B_{1/a}(at)-\lambda mat)/\sqrt{a}\) and the bivariate process \(Z_{a}(t)=(T_{1/a}(at),L_{a}(t)),t\geq 0\). The finite-dimensional distributions of \(Z_{a}(t)\) converge as \(a\rightarrow \infty \) to a process \(V_{l}(t)\), where \(l\) is one of the values \(1,\delta ,\delta ^{2}\) and \(V_{l}(t)\) is known (Theorem 3.1). The proof of the results is based an a detailed use of characteristic functions.