Comparing Differences Across Groups

Abstract

Imagine you want to find out whether people who read statistics books are better at analyzing data than those who do not. You could study this question in at least three ways: (1) You set up an experiment with two groups, one of which you make read a statistics book (a real cruelty), and then you make both groups analyze the same data. (2) You run an experiment with one group only, test their analysis skills, make them read a statistics book, and then test their skills again. (3) You find some people who read statistics books and other people who do not and compare their analysis skills by, for example, studying the number of their quantitative research publications or their grades in statistics classes. All three ways would end up with one variable that tells you whether a person reads statistics books or not—a dichotomous variable that defines group membership—and one continuous variable that summarizes people’s analysis skills (or statistics performance). Answering your research question would require you to evaluate whether the analysis skills of the group that read the book are better than those of the other group. This form of group comparisons—comparing one variable score between two groups—is the simplest. This chapter starts from this simple example and adds complexity by adding more groups and variables of interest.