Null-Hypothesis Testing with Graphs

Abstract

Because biological processes are full of variations, statistics will give no certainties only chances. What chances? Chances that hypotheses are true/untrue. What hypotheses? For example:

• 1. our mean effect is not different from a 0 effect,

• 2. it is really different from a 0 effect,

• 3. it is worse than a 0 effect,

where 0 effect means that your new treatment or any other intervention does not work. Statistics is about estimating such chances/testing such hypotheses. Please note that trials often calculate differences between a test treatment and a control treatment, and, subsequently, test whether this difference is larger than 0. A simple way to reduce a study of two groups of data, and, thus, two means to a single mean and single distribution of data, is to take the difference between the two means and compare it with 0. In Chap. 2 we explained that the data of a trial can be described in the form of a normal distribution graph with SEMs on the x-axis, and that this method is adequate for testing various statistical hypotheses. We will now focus on a very important hypothesis, the null-hypothesis. We will try and make a graph of this null-hypothesis, and then assess whether our result is significantly different from the null-hypothesis.