Journal on Policy and Complex Systems Table 1 . Input Parameters for the Agent-Based Model
puts a set of aggregate statistics at each tick of the simulation , including the count of students by the three potential mismatch values ( 0 , 1 , 2 ), the total number of disruptors , n d
, and the count of students by each class level , TrueL and AssignedL .
Realized GPA . The GPA of the students resulting from the predictive model , decision model , and classroom model is dependent on the true potential GPA of the student , the number of disruptors in the school , and the level of mismatch in the classrooms . The latter two variables are additional hierarchical considerations in the application at the classroom and school levels . These are not guaranteed to be in training data if the focus of the training data is on individual level performance . Variables such as the proliferation of the number of disruptors in the school is completely application dependent and contextually dependent on the predictive model outcomes . In this stylized example , calculation of individual i ’ s realized GPA is a function of both individual and group-level effects ( e . g ., peer influence ) captured only in the complex system model
174