ABSTRACT
Interaction occurs when one factor’s effect on the dependent variable shows a different pattern at the various levels of the other factor. New example is introduced which will serve to illustrate all aspects of the two-factor design. Consider the simple effect of the training factor at the first level of the machine factor. The training factor is seen as having a meaning critically dependent on subjects’ exposure to another factor: type of machine. The definition is put into effect by removing the effects of first one factor and then the other to leave an array of means whose deviations can only represent the effect of interaction. There are several independent estimates of the between-subjects variance arising respectively from the effects of the factors and their interaction. These are unbiased estimates only if the apparent effects of the factors and interactions are caused by sampling fluctuation.
