← Back to projects

COVID-19 Impact on Student Behavior

A statistical study of how lockdown reshaped student life. Using 1,182 survey responses, I modeled the relationships between time-allocation patterns, digital platform use, and health outcomes to identify which behaviors actually protected student wellbeing and learning satisfaction during the pandemic.

The work paired careful data-quality remediation with an inferential toolkit of chi-square tests, ANOVA, and logistic and ordinal regression, so that each reported effect could be defended, not just observed.

Timeline: Jan 2025 – Apr 2025
Role: Statistical Analyst
Stack: R, tidyverse, ggplot2, MASS (ordinal regression)
Dataset: 1,182 student survey responses

Approach

Audited the raw survey export for missing values and categorical inconsistencies (inconsistent labels, free text in coded fields, implausible values), and documented every remediation decision before analysis began.
Used chi-square tests of independence to screen categorical associations between device type, activity patterns, and self-reported outcomes.
Applied ANOVA to compare mean time allocation across groups, checking normality and homogeneity-of-variance assumptions rather than assuming them.
Fit logistic regression for binary outcomes (weight change) and ordinal regression for Likert-scale outcomes (learning satisfaction, digital engagement), keeping the model family matched to the response type.
Built supporting visualizations in ggplot2 so the findings could be read by an educational-planning audience without a statistics background.

Findings

Physical activity was the critical protective factor for health. Time spent on physical activity was the strongest predictor of favorable weight-change outcomes, holding other lockdown behaviors constant.
Device type shaped engagement. Tablet users reported meaningfully higher digital engagement than smartphone or laptop users, a result with direct implications for how remote learning materials are designed and distributed.
Time-allocation patterns during lockdown predicted learning satisfaction, pointing to structured daily routines rather than raw screen time as the lever worth pulling.
Delivered findings as a written report with supporting visualizations aimed at educational planning decisions.