Statistical tests are performed to determine whether an observeddifference is "significant" or due to chance factors.

Think back to the children's game of comparing two drawings. At firstglance, the drawings might look identical. The closer you look, however, themore differences you can see. If you could look at very fine details, you mightsee differences in the color of the ink or the grain of the paper.

Levels of significance are used in statistical tests to let the reader knowhow "close" the statistician "looked" at the two statisticsbeing compared. Levels of significance are also called "p values"because they express the probability that the observed differences are due tochance.

In the children's game, looking closely would allow you to see moredifferences. However, if you looked at the two drawings under a powerfulmicroscope, you would see so much detail that you might lose the big picture. At this level, the drawings might look identical since they are made out of thesame material. The mistake of looking too closely happens often in statistics. Experimenters may choose a level of significance that is too small, and they maynot know it until they perform a significance test.