8 Hypotheses Testing

In Chapter 7 we learned how to look for associations between two variables in random sample data. Just because two variables’ observations exhibit a pattern that we can see in the sample doesn’t mean that the variables are necessarily truly related in the population. Recall the purpose of sampling from Chapter 6: to infer something about a population based on a sample, i.e., to use sample statistics to estimate population parameters.

Given this, the questions you should be asking at this point are: Is an association we observe in the sample data something that exists in the population of interest? That is, do we observe this association because it really exists in the population and is reflected in the sample? Or is our sample unusual enough so that the association is an artifact of random chance, present only in this one sample?  How certain can we be in our conclusion either way?

To answer these questions, you need to learn how to test potential associations for statistical significance. The last section of this chapter and the next two chapters are devoted to just that. First, however, there is some preliminary work to do. To that effect, in this chapter I introduce you to the concept of a hypothesis in social science research and the logic of hypotheses testing, both as a theory and in practical terms.

Before we delve into this (rather extensive) topic, I still have to address the elephant in the room when it comes to statistical associations: causality, next.