This paper describes a 100 x (1 - [alpha]) percent confidence interval for the mean of a bounded random variable that holds for every sample size n and avoids the error of approximation that assuming normality induces.

First order asymptotic validity is established for stationary bootstrapping of the realized volatility. This enables us to construct a bootstrapping confidence interval for integrated volatility. A Monte-Carlo experiment shows that stationary bootstrapping confidence interval is also valid in a...

Casella, Hwang and Robert, Statistica Sinica, 1993, consider a loss function that is a linear combination of the interval length and the indicator function that this interval includes the parameter of interest. They show that this leads to a confidence interval for the normal mean with...

In this work, the exact finite-sample distribution of the median absolute deviation about the median (MAD) of continuous random variables is derived. The main results are used for providing exact confidence intervals for the scale parameter of a location–scale model, which are robust to the...