How to explain an interaction in statistics? During last spring semester, I’ve come up with an example: tea sweetening.
Let’s consider two factors: adding sugar and stirring. The dependent variable is the taste of tea (bitter or sweet). First, let’s consider stirring alone, without adding sugar. Well, it doesn’t matter how much of stirring you do, the tea will remain as bitter as it was at the beginning. Conclusion: stirring has no effect on the taste of tea.
Second factor, adding sugar alone, without stirring. You can add two or three teaspoons of sugar, without being able to sense any difference. After adding a lot of teaspoons, you start feeling that it’s becoming sweet. Conclusion: adding sugar has a very weak influence on the taste of tea.
Now, the third experiment, adding sugar and stirring, together. Wow, it looks like it’s enough to add just one teaspoon of sugar and stir a little and you can already feel the difference.
Conclusion: There is an interaction between adding sugar and stirring.
It’s not the sugar or the stirring alone that makes the tea sweet. It’s the fact that those two things happen together.