Strong limit theorems of empirical distributions for large segmental exceedances of partial sums of Markov variables (Q1180583): Difference between revisions

In the paper reviewed above the authors characterized the composition of high scoring segments among partial sums of i.i.d. random variables. In this paper the authors consider a letter sequence \(A_ 1,\dots ,A_ n\) assuming values from a finite alphabet \(\{ a_ i\} _{i=1,\dots ,r}\) governed by an \(r\)-state irreducible Markov chain. Suppose \((X_ m,U_ m)\) are independently distributed given the sequence \(A_ 1,A_ 2,\dots \), where the joint distribution of \((X_ m,U_ m)\) depends only on \(A_{m-1}\) and \(A_ m\) and is of bounded support. When \(A_ 0\) is started with its stationary distribution, \(E\;X_ 1<0\) and the existence of a finite cycle \(C=\{ A_ 0=i_ 0,\dots ,A_ k -i_ k =i_ 0\} \) such that \(P\bigl\{ S_ m = \sum _{i=1}^ m X_ i >0,\;m = 1,\dots ,k;C\bigr\} >0\) is assumed. Define the stopping times \[ K_ 0=0,\quad K_ \nu = \min \bigl\{ k\geq K_{\nu-1} +1,\;S_ k - S_{K_{\nu-1}} \leq 0\bigr\} ,\quad \nu\geq 1, \] and for \(y>0\), \[ T_ \nu (g) = \min \bigl\{ m:\;m>K_{\nu -1}\text{ and either }S_ m - S_{K_{\nu - 1}}\leq 0\text{ or }S_ m - S_{K_{\nu-1}} \geq y\bigr\}, \] \[ L_ \nu (y)= T_ \nu (g) - K_{\nu -1},\;W_ \nu (g) = \sum_{m=K_{\nu -1}+1}^{T_ \nu (g)} U_ m . \] Theorems 1 and 2: Let \(I_ \nu (y)=1\) while \(I_ 1 (y) = \cdots = I_{\nu -1} (y) = 0\). Then \(L_ \nu (y)/y \to 1/w^*\) a.s. as \(t\to \infty\), and \(W_ \nu (y) / L_ \nu (g) \to u^*\) a.s. as \(y \to \infty\), for any value of \(A_ 0\), certain positive constants \(w^*\), \(u^*\).