Abstract
The robustness to a single outlier of four methods for estimating the mean is investigated. The methods appear in such officially published and widely used standards as those of the International Organization for Standardization and Japanese Industrial Standards. Using a computer simulation, we compare the bias (the expected difference between the estimate produced and the true value), the variance of the estimates, and the average sample number required, for each of the four methods.
Additional information
Notes on contributors
Yoshikazu Ojima
Dr. Ojima is on the Faculty of Engineering.