Pseudo-maximization and self-normalized processes

Abstract

Self-normalized processes are basic to many probabilistic and statistical studies. They arise naturally in the the study of stochastic inte- grals, martingale inequalities and limit theorems, likelihood-based methods in hypothesistesting and parameterestimation, and Studentizedpivots and bootstrap-t methods for confidence intervals. In contrast to standard nor- malization, large values of the observationsplay a lesser role as they appear both in the numerator and its self-normalized denominator, thereby mak- ing the process scale invariantand contributing to its robustness. Herein we survey a number of results for self-normalized processes in the case of de- pendentvariablesand describe a key method called "pseudo-maximization" that has been used to derive these results. In the multivariate case, self- normalization consists of multiplying by the inverse of a positive definite matrix (instead of dividing by a positive random variable as in the scalar case) and is ubiquitous in statistical applications, examples of which are given. AMS 2000 subject classifications: Primary 60K35, 60K35; secondary 60K35.