Multilevel Models in Humanities; Case Study of Nationwide Exam Participants

In many scientific studies, especially in the humanities, the influence of variables on each other, or prediction of one variable by some other variables is of special interest. In statistical problems, these goals can be achieved by developing proper statistical models. In recent decades, among various statistical models used by humanities scholars, multilevel models have got special attention. In these models, instead of assuming that model parameters are constant, the impacts of interaction between high levels and low levels have been widely seen in human sciences research. In addition, these models in comparison with low-level models, modify the bias of estimated parameters resulted from cluster sampling. This paper presents multilevel models, and addresses their applications, advantages and disadvantages. Moreover, a three-level model was used to analyze the role of constituent variables of socioeconomic status in educational performance (total score) to enter into universities and “higher education institutions” over time. In this example, the effect of each constituent variable, including parent's educational level, father's occupation, level of income in different genders, was investigated considering the status of the city and province from where participants in the experimental science group of the entrance exam of higher education came during the years 2001 to 2009.