Normative Data Accounting for a Binary Independent Variable

This chapter details how regression-based normative data can be derived in a setting where a binary independent variable (i.e., a variable with only two possible outcome levels, such as Gender) has to be accounted for in the norms. Some important concepts of regression analysis that are relevant in a normative data context are discussed, such as the sampling distribution of an estimated parameter and inferential procedures for the mean structure of the model. It is further explained that there are two distinct stages in the derivation of regression-based norms. First, a parsimonious regression model is built for the mean and the variance structures in the normative dataset. Second, the actual norms are derived in the form of, e.g., a classical paper-based normative table or an automatic scoring program that is implemented in a spreadsheet. The methodology is illustrated based on two case studies in which regression-based normative data that account for Gender are derived for the Taylor Manifest Anxiety Scale score and the Science Exam subscale score of the General Certificate of Secondary Education test. These analyses are conducted using the NormData package in the R software.