A dependent data extension of Wald's identity and its application to sequential test performance computation

The development of an exponential transformation of measure not restricted to independent identically distributed data is presented for a class of semi-Markov processes known as Markov-additive (MA) process. The MA measure transformation is applied to obtain a new form of the Wald identity. To demonstrate the utility of this extension to MA processes, the application of sequential tests with Markov data to error probability performance analysis is investigated. In particular, it is demonstrated how previous work dealing with robust error probability evaluation in the presence of uncertain serial dependency can be applied to sequential tests.<
>