ABSTRACT
This chapter identifies two types of questions: questions of association where the frequencies of occurrence of different combinations are not predictable in advance; and questions of goodness-of-fit or agreement where one can use independent information to predict expected frequencies. It analyses questions of association using contingency table χ2; questions of goodness-of-fit are analysed by the goodness-of-fit χ2. The chapter outlines various constraints to the use of χ2. When statistically significant results are obtained in questions of association, the chapter uses standardized residuals to identify significant associations. The chapter also uses various methods dependent upon the biological and statistical context of the problem to follow up deviations from predictions in the goodness-of-fit χ2. This modification to the method is required because the χ2 distribution is a continuous distribution that we are using with discrete (frequency) data.
