There is a three-level structure in school: students are nested within classrooms (or teachers), and classrooms are nested within schools. The study implemented the generalizability theory to decompose the total variance, so that the reliability and standard error of mean at school level can be obtained. However, from the definition of G coefficient, ignoring the classroom-level will lead to different estimates. In addition, specifying the classroom-level as fixed or random effect influences the results as well. Figure 1 summarized the performance of G coefficients when classroom-level is fixed, random, or ignored. Similarly, Figure 2 summarized the standard errors of school means under the same conditions.