ABSTRACT
If the dependent variable is a dichotomous variable or a proportion, both the assumption of continuous scores and the normality assumption are violated. This chapter treats multilevel regression models for such data. These models are based on generalized linear regression. This chapter describes multilevel generalized regression analysis and two different approaches to estimation: Taylor series linearization and numerical approximation. If available, numerical approximation is preferred because it is more accurate. Next, this chapter treats standardization and explained variance in these models. It ends with a discussion of interpretation issues.
