Student Mental Health Dashboard with R Shiny
https://github.com/robertjgarciaphd/Student-Mental-Health-R-Shiny.git
Overview
As someone with six years of teaching experience at the university level, I have firsthand familiarity with a wide range of student mental health concerns and their impact on academic performance. I was curious to explore this from a research and data analytics perspective. For this project, I chose to examine the research question of how student mental health varies across demographics. To answer this question, I chose to create an app that:
- Visualizes data
- Allows users to apply filters
- Allows users to adjust the display in real-time
This dataset consisted of N = 101 students at the International Islamic University of Malaysia. This dataset was hosted on Kaggle.
Columns in the data set included:
I divided these columns into two categories:
- Predictor variables
- Outcome variables
App Design
Key considerations for my app design included:
- Allowing users to filter by student demographics
- Default setting: Include all students in the sample
- Showing how many students are left after filtering
- Showing all output/dependent variables simultaneously
App Interface
Insights
The sample size was not just small, but it turned out to be highly skewed as well. Only 8 students in the sample were in Year 4, making comparisons across students' year of study impossible. Less surprisingly, there were not enough students in each of the numerous majors to make comparisons. The skew extended to the variables of Gender (74.3%), CGPA (85+% with 3.00 cumulative GPA or higher), and marital status (84.2% unmarried). In short, all the input variables were skewed, making it difficult to make any conclusions, especially given the small sample size.
Practical Value
Despite the lack of insight we can extract from the data, this project still was useful as a proof of concept for mental health dashboards. It provides easy snapshots of mental health outcomes like depression, anxiety, and panic attacks to help identify which student groups are most at risk. This could be shared with counselors and administrators to help allocate mental health services and resources. It could also be adapted to corporate contexts to keep tabs on employee well-being with different outcome variables.
Future Directions
Some considerations to expand upon this work include uploading a larger data set with less skew and more students in each major. This would enable us to make statistically-informed conclusions. It would also be helpful to have longitudinal data to track the progress of treatment and understand which students are or are not improving.