Statistical models for the randomness in data usually involve one or more unknown parameters. The values of these parameters are often of particular interest and we have described already described some methods to estimate them. With estimation, we try to answer questions of the form "what is the value of a parameter" — questions that hope for a number as the answer.

Another branch of statistical inference is called hypothesis testing. It tries to assess whether some claim about the parameter values is true, and hopes for an answer of "yes" or "no".

As with estimation, hypothesis tests can rarely give a perfect answer to the question. The data that have been collected only provide limited information about the values of the parameters, so there must be some uncertainty expressed in the conclusions. In estimation, this uncertainty is usually expressed with a confidence interval. This chapter describes how hypothesis tests should be performed and how the uncertainty in their answers should be expressed.