Predictive model of mathematical literacy ability, specifically university students' critical thinking skills: A case study on the inappropriateness of multilevel regression

Abstract

This study aims to analyze the influence of Project-Based Learning (PjBL) and Guided Discovery Learning (GDL) on students' critical thinking skills in geometry focusing on mathematical literacy within geometry content. Specifically, it examines the effects of self-confidence, mathematical literacy scores, and gender on critical thinking skills and compares the predictive accuracy of multiple linear regression and multilevel regression models. A quantitative approach with regression analysis was used. Multiple linear regression and multilevel regression models were applied to assess the relationships among variables. Model accuracy was evaluated using intraclass correlation coefficients (ICC) to determine the impact of class-level grouping. Self-confidence and mathematical literacy in geometry significantly influenced students' critical thinking skills. Gender showed no significant effect. The multiple linear regression model outperformed the multilevel model as reflected by lower ICC values, indicating that class grouping did not substantially affect critical thinking outcomes. Internal factors, such as confidence and subject mastery  are more critical to students’ critical thinking development than external factors like gender or class grouping. Teachers should prioritize enhancing students’ confidence and content mastery through well-designed instructional strategies to foster critical thinking in mathematics learning.