A new statistical method for analyzing children’s longitudinal cognitive development based on a measurement invariance condition

Cognition is not only the core of human intelligence but also an essential ability of children. In previous studies, longitudinal methods have frequently been used for analyzing children’s cognitive development. However, few researchers have considered the measurement invariance problem in longitudinal studies. Thus, their results may be biased or even misleading. The purpose of our study is first, to use a newly developed MIRT method to analyze the longitudinal trend of preschool children’s cognitive development; second, to analyze whether the instrument satisfied the measurement invariance condition or not; third to find items that did not satisfy measurement invariance, to help future revisions of the instrument. In this study, we chose 882 48-month-old children from the country and tested their cognitive development in the second year. After model comparison and considering the instrument construct, we choose a 2-tier item response model to analyze the instrument and to test the test-level measurement invariance and item-level measurement invariance. The result show that the instrument is partially invariant, which shows that the construct remains stable over age. On the other hand, items 2 and 3 did not satisfy the measurement invariance condition. Thus researchers can revise or delete these items to ensure the quality of the instrument. Moreover, the children’s cognitive proficiency grows significantly from age 4 to 5.