Maybe you have seen the children's game composed of two similar drawingsside by side. Players compare the two drawings to determine if and how they aredifferent. Children do this simply by looking closely at both drawings.
When statisticians compare two statistics, they need special tests todetermine if there is a significantdifference between them. These tests are called significance tests orhypothesis tests.
Jodie is researching the effect of urban life on children. She hasperformed the same study of memory with two groups of children, one rural andone urban. Jodie wants to see if there is a significant difference between theperformances of the two groups of children. To do this, Jodie needs to do asignificance test.
The first thing Jodie must do is form a nullhypothesis. The null hypothesis always states that there is no differencebetween thetwo groups being tested. So, in our example, Jodie's null hypothesis is: There is no real difference betweenthe rural group and the urban group's performance in the memory study.
She then assigns a level of significanceto her test. She chooses .01 as her p-value.
Jodie will now test her null hypothesis at the .01 level of significance. She will perform a t-test (one kind ofsignificance test) and will decide if she can REJECT her null hypothesis.
"Rejecting her null hypothesis" is the same as saying the oppositeof her null hypothesis: There IS a difference between the rural group and theurban group. If Jodie can reject her null hypothesis, she may conclude thaturban living impacts memory of children.
If Jodie cannot reject her null hypothesis, she has a choice. She could:
Accepting the null hypothesis could be an error. There may be a realdifference that is too small to be detected by her first test.