Demands and expectations for data science education have been increasing in recent years. Data science education requires not just acquiring knowledge and skills, but also the ability to use them to solve real-world problems. To achieve this, it is recommended that project-based learning be conducted using realistic data. In this paper we reported on a PBL-type teaching practice carried out in collaboration among three universities. The classes were structured to allow communication between students with different knowledge and experience in data science education and between students with different fields of expertise. In addition, we provided opportunities for graduate students to answer questions from undergraduate students in order to develop consulting skills within the class. By analyzing real data, the students became more interested in data science and actively engaged in problem solving. Regarding consulting, the questioner was asked to confirm the knowledge he or she lacked in solving the problem, while the respondent made an effort to explain things in a way that suited the questioner. This paper provides an overview of the classes conducted collaboratively by the three universities and discusses the effects of these collaborative classes on student learning.
