Dissertations and research projects

Hypothesis testing

A hypothesis is a predictive statement that can be tested through the collection of data. The data can be analysed and can either provide support for, or help to reject, a hypothesis; this in turn should allow a researcher to draw some conclusions about what they are investigating.

Null and alternative hypotheses

Hypothesis are classified by the way they describe the expected association/difference between variables. When we test our hypothesis/hypotheses it is important to remember we are testing it against the assumption that there isn’t an association/difference between the independent and dependent variables: we call this the null hypothesis. By testing this assumption, statistical tests can estimate how likely it is that any observed association/difference between variables is due to chance.

In addition to the null hypothesis we also have the alternative hypothesis. This hypothesis states that there is an association/difference between groups; this cannot be tested directly but can be accepted by rejecting the null hypothesis. This is achieved through statistical tests that can help to demonstrate that any observed association/differences are not due to chance. Once this is established, we can accept our alternative hypothesis and start to draw conclusions from our data.

Hypotheses can either be one-tailed or two-tailed:

• One-tailed hypothesis –specifies the direction of the predicted association between the independent and dependent variable. For example, the higher an individual’s educational level, the more books they will read in a one-year period.
• Two-tailed hypothesis – does not specify the direction of the predicted association between variables; only that an association exists. For example, there will a be difference in the number of books read in a one-year period, dependent on the level of an individual’s education.

Key things to remember when writing your hypothesis/hypotheses:

• Your hypothesis should always be written as a statement and before any data are collected.
• It should be simple and specific; include the variables, using concise operational definitions, and the predicted relationship between these variables. If you have several predictor (independent) variables it would be better to write several simple hypotheses – think one predictor and one outcome variable.
• Always keep your language clear and focused.