The purpose of this study was to propose a mixed-methods approach to analyzing shared epistemic agency in jigsaw instruction from multiple temporal perspectives, and to evaluate its effectiveness by examining actual datasets. We employed a combination of socio-semantic network analysis (SSNA) and in-depth dialogical discourse analysis as a mixed-methods approach, and analyzed discourse transcripts by university students engaged in jigsaw instruction. First, we graphically depicted a quantitative measure of shared epistemic agency at the group level and identified pivotal points of discourse where students might engage in an epistemic action toward alleviating a lack of knowledge. Then, we conducted dialogical discourse analysis of the segments around the pivotal points to describe students' collaboration practices. SSNA represented the quantitative nature of shared epistemic agency with 60% accuracy and provided a new way to look at it as a distribution of pivotal points for alleviating a lack of knowledge across all processes of jigsaw group activities. The dialogical discourse analysis of the discourse segments identified by SSNA further described dialogical patterns in the shared epistemic agency and each student's contribution to them.