Analysis of measurement of metacognitive skills and students' self-efficacy using the Rasch model

Abstract

This study aimed to develop an instrument to measure students' metacognitive skills and self-confidence, focusing on Statistics and Data Science courses at a Police Science College in Jakarta, Indonesia. Using a quantitative approach, the population included all students enrolled in the 2024–2025 academic year from three study programs. A total of 142 students were selected through random group sampling, with 45 students representing each program. Data were collected via a Google Form containing 25 multiple-choice items for metacognitive assessment and 25 Likert-scale statements for self-confidence measurement. Data analysis used Winstep, Jamovi, and ConQuest software, based on the Rasch model's assumptions: unidimensionality, local independence, parameter invariance, and log-linear properties. The findings showed that 73.7% of participants had intermediate to high ability levels, with raw scores between 21–24. No extreme scores (0 or 25) were found, suggesting a relatively homogeneous ability distribution. In contrast, self-confidence levels were higher across the sample. The combined average logit score for metacognitive skills and self-confidence was +3.08, indicating high performance and confidence among students. These results support the reliability of the developed instrument and provide insight into students’ cognitive and affective readiness in quantitative courses within police science education.