Introduction to measures of association

Abstract

We have previously made the general point that statistical significance and theoretical or practical significance are not the same. We may, for example, find a relationship between age and exam grades. However, the mere fact that there is a difference in exam scores between age groups gives no guidance as to whether the relationship between these variables is very strong or very weak. Should it warrant running separate classes for students in different age groups? Does it justify a change in the current thinking about education and cognitive skills? Tests of significance do not help much in answering such questions. They do not measure the strength of a relationship, but rather indicate whether a relationship between two variables exists in the population. Measures of association, on the other hand, are designed to provide such information.