Confidence interval determination of on-time performance within measurement of the transit time for single piece priority mail

Abstract

The paper is focused on usage of methods for confidence interval calculation by measurement of the transit time of end-to-end services for single piece priority mail. The results come out from modelling of measurement system on the basis of relevant field of study with particular discriminant characteristics. The role of confidence interval is the expression of estimate accuracy for on-time probability. The accuracy is expressed as width of the confidence interval of on-time probability estimate. The probability distribution for modelling of on-time performance in simple random sample is based on relevant approximation of binomial distribution. Considered approximations are normal distribution, Agresti-Coull approximation and approximation by beta inverse function. The main objective of this paper is calculation of confidence interval by mentioned approximation methods, comparison and assessment of the results from viewpoint of next modelling. The results of on-time probability estimate accuracy are related to input assumptions, i.e. especially geographical coverage of postal services and geographical stratification on disjunctive set of postal areas. Calculation methods consider one-Operator field of study with domestic mail for various sample sizes.