Applying multilevel regression to ordered response variables: an analysis of trust in the prosecutor's office

초록

The multilevel regression model is primarily employed in analyzing data with a hierarchical structure. However, existing literature has rarely addressed the challenges of applying the multilevel regression to an ordered categorical response variable. This study aims to explore the multilevel ordered logistic regression model, with a focus on sample size and model estimation. It reviews the theoretical foundations of the model and conducts a comparative analysis of various estimators through simulation. Specifically, the study compares maximum likelihood (ML), penalized quasi-likelihood (PQL), and Bayesian estimators, especially in scenarios with smaller sample sizes at higher group levels. Additionally, this study analyzes real data to identify factors influencing trust in the prosecutor's office in Korea.

키워드