The presence of interaction can have important implications for the interpretation of statistical models. If two variables of interest interact, the relationship between each of the interacting variables and a third “dependent variable” depends on the value of the other interacting variables and this makes it hardest to anticipate or predict the consequences of the value of variable that changes particularly if the variable it interacts with are difficult to control. (Eastern & McColl 2016)
Example is if we want to examine the effect of two variables, gender and premature birth on health outcomes, we would first of all outline any differences in health outcome score among gender as a main effect. Similarly, we will describe any difference in the scores of full term/premature as a main effect. The presence of an interaction effect shows that the effect of gender on health outcome varies as a function of premature birth status.
Easton J.C & McColl 2016: Design of Experiments and Anova. Retrieved August 17, 2018 from https://www.stats.gla.ac.uk/steps/glossar/anova.html#intermpediaiew.com.