In many practical applications, one is interested in obtaining confidence intervals for nonlinear functions of the parameters. This paper considers the following different methods: Fieller's method, Taylor's series expansion, and bootstrap methods. Compared to some of the earlier results in the empirical studies that are against the application of bootstrap, our results suggest a different conclusion in favor of the bootstrap methods.
This content is only available as a PDF.
© 1999 President and Fellows of Harvard College and the Massachusetts Institute of Technology