Inference on rare errors using asymptotic expansions and bootstrap calibration

The number of items in error in an audit population is usually quite small, whereas the error distribution is typically highly skewed to the right. For applications in statistical auditing, where line item sampling is appropriate, a new upper confidence limit for the total error amount in an audit population is obtained. Our method involves an empirical Cornish-Fisher expansion in the first place; in the second stage we employ the bootstrap to calibrate the coverage probability of the resulting interval estimate.