ABSTRACT
This chapter starts by a generalization of the bootstrap distribution. It examines some material necessary for the comparison of asymptotic properties and for the development of a more theoretical view of the bootstrap ideas. The chapter discusses some more specific confidence interval methods for independent and identically distributed data. It explains the identical distributions by considering regression problems, and examines independence by considering some methodology for stochastic processes. Many problems also have good classical solutions when a parametric likelihood can be written down. The practical need for this kind of bootstrap is therefore perhaps somewhat limited by these facts. A root variable with unknown distribution can be transformed into an approximate pivot variable by a bootstrap method. Defining the new variable as a new root, another outer bootstrap analysis estimates the distribution of this root function. The outstanding bootstrap contribution is to give uncertainties of the estimated parameters also in very complicated analyses.
