Strong approximations of semimartingales by processes with independent increments (Q2640221)

This paper provides some very general strong approximations for continuous time semimartingales by continuous processes with independent increments whose variance functions increase polynomially or exponentially. The basic assumptions involve constraints on the probability of big jumps and on the differences between the time scales and trends of a suitably truncated local martingale part of the process and those of the approximating process. No moment assumptions are required. Applications are given to provide an almost sure invariance principle for partial sum processes whose variance grows nonlinearly and also one for the functionals of a Markov process.