Learning analytics in Human Histology reveals different student'clusters and different academic performance

Resumen

Universities and Higher Education institutions have created different platforms that provide digital environments with private access. These digital spaces simulate physical spaces for teaching and learning, allowing interaction between participants. The interactions are stored in the platforms so data can be analyzed to reveal behaviors and preferences. The literature shows different patterns in student's behavior and it has been demonstrated that different clusters obtain different academic performances, as well as the importance of the virtual spaces. Human Histology is a second-year compulsory subject of the Medicine Degree at the Universidad Complutense de Madrid. It lays the foundation for Pathology learning in the third-year. Practice classes are focused on the observation of histological slides so students learn general and differential microscopic characteristics of organs. Students are expected to identify each organ under the microscope. Some years ago, teachers decided to make the evaluation a continuous process through minitests, short tests with projected images, and a teamwork consisting in making a notebook with freehand drawings showing the different organs and histological staining procedures. The final practice mark was obtained as the addition of the continuous activities (minitests 20% and teamwork 25%), the final exam (45%), and class attendance (10%). In the virtual space created for managing Histology practices, different resources were offered such as scripts for each session, histological images file and URLs that link to histological atlas and other web sites. We present in this paper the results obtained when processing the logs of the virtual space with a free software environment for statistical computing and graphics named RStudio, an integrated development environment for R. A total of 25583 logs corresponding to the registered activity in course 2018/19 were refined and subsequently analyzed. The quantitative measures chosen were the number of total logins in the virtualized course per day, the average of the login frequency per each day of the week and per each hour of the day, the number of entries in resources per day and the number of entries in URLs per day. Also, a statistical analysis of the data was performed with SPSS 25 software, comparing the use of virtual campus to academic performance. Non-parametric Spearman correlation tests and decision trees with two cut criteria were obtained. Results show that the activity in virtual campus is clearly conditioned by due dates (dates of minitests and final exam and deadline to submit the teamwork). Decision trees reveal different clusters of students according to the variables number of visits, number of entries to resources and to URLs, and that these clusters get different marks in both the minitests and the final exam in addition to the final mark.