ABSTRACT
This chapter introduces the readers to a series of log-linear models for rater agreement. They proceed in three steps. First, they introduce a base model. Second, they introduce a more general manifest variable model of rater agreement. Third, they introduce a selection of specific models. The design matrix approach is widely used in analysis of variance and regression analysis, and has also been employed in log-linear modeling. The resulting model is called equal-weight agreement model or diagonal set model. The model considers each rating category equally important for the assessment of rater agreement. The Equal-Weight Agreement Model requires the assumption that agreement is equally important or carries equal weights across all categories. The term covariate is often used simply as an alternative name for explanatory variables that are likely to affect the response variable. Students are often suspected of gravitating toward those examiners that are believed to exhibit such trends as leniency.
