8c66bb19847dd8c21413c5c8c9d68306-AuthorFeedback.pdf

Please see below for our response. Reviewer 2: Regarding the baseline methods in the experiments: Original LiNGAM assumes that there is no confounder. So the issue is that it is not clear how to compare its result with the groundtruth graph (with confounders). For clustering methods, clearly, different assumptions for clustering will lead to different clustering results. Reviewer 3: 1. Regarding "the setting is limited": We totally agree, and at the same time would like to mention Theorem 1 and Proposition 2 ensure the correctness of Phase 2 of our method (Algorithm 2).