| Abstract: | This paper reviews what is currently known about the behaviour of the t-statistic when one is no longer sampling from a normal distribution. Suppose Y is a batch of data on which the t-test is performed. Briefly then, heavy-tailed components of Y give a light-tailed t , positive correlation among Y gives a heavy-tailed t, and positively skewed components of Y give a negatively skewed t. The emphasis is on understanding why one gets this type of behaviour, although some numerical tables are presented to illustrate the conclusions. |
