This paper describes methodological considerations for fusing data sources in understanding both affective and problem solving states of students as they engage in computational thinking (CT) game-based learning. We provide both a theoretical and empirical rationale for using data including facial recognition and students’ logfile data to gain a more robust explanation of why students may experience emotions such as frustration during CT game-based learning activities. We showcase illustrative examples using data from individual learners to highlight the methodological approaches that we have taken. Finally, given the complexities of understanding constructs such as affect and problem solving, we provide a rationale for using a data fusion methodological approach.