An Investigation of the Robustness of Student'S T Test Using beta and Erland Distributions

摘要

The robustness of 'Student's' t-test with regard to the assumption of normality is investigated. This is accomplished by empirically developing the distribution of a new statistic from a large number of samples of varying sizes from the Beta and the gamma distributions using different values of the parameters. The various significance levels of this new distribution are then compared with the corresponding significance levels from 'Student's' t distribution using an IBM 360 computer. A comparison of the distribution frequencies with the standard normal distribution frequencies is also presented. (Author)