Pilot study on the validity and reliability of MIM: An alternative assessment for measuring metacognition in mathematics among college students

Metacognition could measure students' true mathematical abilities. However, a domain-specific instrument in measuring the metacognition in mathematics is very limited in the field. In this study, a reliable and valid instrument called “MIM” is proposed. It measures all components in metacognition according to Brown's model. 1100 senior one students of a public secondary school are enrolled in stage A while 14 senior two and 14 senior three voluntary students studying an international program are enrolled in stage B in this research. The reliability of the MIM is examined (Cronbach α = .905) while the Spilt-half reliability test is done (Cronbach α = .825). The factor analysis with Oblimin rotation yielded four factors which were classified as ‘prediction’, ‘planning’, ‘monitoring’ and ‘evaluation’ according to the content of the items. Meanwhile, the MIM yielded a small linear relationship with the HSEE (r =.285, p < .001) which is highly consistent with recent research findings. Planning seems to have the highest linear relationship with mathematics performance among all metacognitive skills (r =.346, p < .001). The MIM has been proven to be valid instrument although it is a self-rated instrument. Its self-rated items are found to have significant relationship with the peer-rated items (r =.509, p = .008), On the other hand, the academic performance of senior two students (M = 76, SD = 22.2) was found to be significantly lower than those of senior three students (M = 170, SD = 19.0) while no significant difference was found in metacognition. It is believed that the CSK would become more and more dominant in predicting the academic success when the level of study increases. Moreover, more variance of the HSEE could be explained by a quadratic (R² = .43, F(2,1096) = 24.7, p < .001) or a cubic model (R² = .45, F(3,1095) = 17.1, p < .001) than the linear model (R² = .37,F(l,1097) = 42.3,p < .001). It is believed that the proficiency of the monitoring skill is one of the main reasons.