Student engagement is a reflection of active involvement in learning. In digital learning environment, research studies on engagement have been focused on detecting behavioral and psychological engagement indicators from the patterns of activities using feature engineering, but student engagement estimates were rarely compared across sessions or across domains of learning. This paper describes how this could be done by revisiting engagement instrument, diagnosing engagement indicators, estimating engagement parameters, and equating. This study illustrates how engagement reliability can be improved by refining engagement indictors. We demonstrated through DataShop data that student engagement levels can be compared across domains of learning.