Moment Inequalities via Martingale Methods
In this chapter we establish different kinds of moment inequalities for partial sums and the maximum of partial sums of a large class of random variables, including martingale sequences, mixingales, and other dependent structures. All the bounds involve the moments of the conditional expectations of either the partial sums or the individual random variables. In most of the proofs martingale approximations are used. This method allows us to use the moment inequalities for the martingale part developed in Chapter 2. We start with a dyadic scheme useful for analysis of the variance of partial sums in the stationary setting. Then, we obtain Burkholder-type inequalities via Maxwell–Woodroofe-type characteristics and an extension of Doob’s maximal inequality for adapted sequences. A Rosenthal-type inequality for stationary sequences is also provided with bounds using conditional expectations of the partial sums. Maximal exponential inequalities are established involving either Maxwell–Woodroofe-type characteristics or the projective operators.