BMO-sequences and amarts (Q5916481)

Let (\(\Omega\),\({\mathcal F},P)\) be a probability space and (\({\mathcal F}_ n)_{n\geq 0}\) be an increasing sequence of sub-\(\sigma\)-fields of \({\mathcal F}\). An adapted sequence of random variables \((X_ n)_{n\geq 0}\) is called a generalized BMO-sequence if \[ \sup \{\| E(| X_ m- X_{n-1}| | {\mathcal F}_ n)\|_{\infty}:\quad 1\leq n\leq m\}<\infty. \] The paper deals with the connection between generalized BMO-sequences and the theory of amarts. The main results are the following: - Every BMO-sequence is a uniform semiamart having the optional sampling property. - Every BMO-sequence for which the net \((X_{\tau})_{\tau \in T}\) converges in probability as \(\tau\) runs over the set T of all bounded stopping times, is an amart. - Every BMO-sequence converging in probability is an amart of finite order. Further, a necessary and sufficient condition is established for an amart to be a BMO-sequence. Generalizations for \({\mathcal K}_ p\) and \({\mathcal K}^+_ p\)-sequences are also presented.