ABSTRACT
This chapter focuses on two methods that are similar to multiple regression—logistic regression and multilevel modelling—but that require specialized analysis with something other than the linear regression procedures in general statistics programs. Logistic regression is useful when the dependent variable students are interested in is dichotomous. Multilevel modelling takes into account the often-nested or clustered nature of our data, such as children within schools, or individuals within families. Logistic regression is most commonly used to predict a dichotomous dependent variable from multiple continuous or continuous and categorical independent variables. Dichotomization is a disappointingly common occurrence in logistic regression research: researchers take perfectly good continuous dependent variables and turn them into dichotomous ones. Multilevel modeling is a regression method that can take into account data that are clustered in some way—students in schools, people in couples, and so on.
