On sensitivity of exponentiality tests to data rounding: a Monte Carlo simulation study

Abstract

Different statistical procedures are differently sensitive to data rounding. It turns out that tests for exponentiality are more sensitive to the data rounding than many classical parametric tests or than nonparametric tests for normality. In this work we find out which exponentiality tests are more robust and which ones are less robust to the rounding. The main tool is Monte Carlo simulation. We estimate and compare the probability of Type I error of nineteen exponentiality tests for different rounding levels and different sample sizes.