【关 键 词 】：
学习投入； 学习分析； 经验取样法； 交叉滞后分析； 纵向研究
In the study of online and blended learning engagement, it is an important challenge for researchers to collect the data of real-time engagement generated in the process of learning, analyze the characteristics of engagement of different groups, and explores the complex relationship between cognition, emotion and affective engagement. This study proposes a dynamic analysis framework of learning input for real-time process data, including immediacy, sustainability and multi-dimension. Based on this, a design idea of longitudinal study on learning engagement is put forward. This study adopts empirical sampling method, cross-lag analysis and cluster analysis to analyze the blended learning engagement. It is found that learners can be grouped into four groups based on their cognitive, emotional and behavioral levels, namely shallow engagement, medium engagement, deep engagement and cheerful engagement, and learners' cognitive, behavioral and emotional engagement is not balanced. The predictive relationship between cognition, emotion and behavioral engagement is likely to be influenced by time, learning environment and other factors. The research results further indicate that the longitudinal study for real-time data acquisition and analysis provides an effective way to accurately describe the characteristics of learners' engagement, and also provides a possibility to reveal the predictive relationship between sub-dimensions of learning engagement and the mediating factors.