|
Abstract:
|
Coverage intervals for a parameter estimated from a complex probability sample are usually constructed by assuming that the parameter estimate has an asymptotically normal distribution, and the measure of the estimator’s variance is roughly chi-squared. The size of the sample and the nature of the parameter being estimated render this conventional “Wald” methodology dubious when constructing coverage intervals, especially for proportions. A revised method of coverage-interval construction has been developed in the literature that “speeds up the asymptotics” by incorporating an estimated skewness measure. We will discuss how skewness-adjusted coverage intervals can be computed in some common situations and why it may be inappropriate to call them “confidence intervals.”
|
