Continuous Predictors 12with Nonindependent Observations One of the distinct advantages of our model comparison approach is that we have been able to integrate the traditionally rather separate analytic frameworks underlying the analysis of designs with categorical independent variables (ANOVA) and the analysis of correlational data, with more or less continuously measured independent variables (multiple regression/correlation). In terms of model estimation and comparison, the model comparison approach is appropriate regardless of the measurement metric of predictor variables. One simply has to learn new ways in which parameter estimates can be interpreted (i.e., as mean diﬀerences or adjusted mean diﬀerences in the case of categorical variables as well as the traditional slope interpretations). This same advantage for our approach extends to the treatment of designs in which observations are nonindependent because they are grouped in some way. As we will show in this chapter, one can readily incorporate more or less continuously varying independent variables into the models underlying the analysis of nonindependent observations.