
Chapter 8
Resampling Methods for Interval
Estimation
Suppose that observations for a random sample are available. The question of interest
is to estimate a statistic
θ
of the underlying distribution. This statistic may be uni-
variate such as mean, variance, median, mode, or bivariate such as correlation coeffi-
cient. If the underlying distribution is known, then we can apply developed theories
to find a sampling distribution and to give a point estimate and confidence interval
in the parametric setting. However, if nothing is known about the distribution, then
nonparametric methods must be employed. We consider two relevant techniques:
jackknife and bootstrap. ...