png

Student Performance Analysis

You can access the full Python notebook here.

1. Introduction

Academic performance is a multifaceted outcome influenced by a dynamic interplay of personal habits, environmental factors, and socio-demographic characteristics. As educational institutions increasingly recognize the importance of holistic student development, understanding how lifestyle choices—such as study routines, sleep patterns, digital consumption, physical activity, and mental well-being—affect learning outcomes becomes critical. This report presents a comprehensive analysis of these relationships using a simulated dataset of 1,000 student records, encompassing over 15 variables spanning behavioral, psychological, and contextual domains.

The primary objective of this study is to identify key predictors of academic success, operationalized through final exam scores, and to quantify their impact. By leveraging rigorous statistical methods—including correlation analyses, t-tests, ANOVA, and non-parametric alternatives—we examine hypotheses related to:

  • The influence of study habits and attendance on performance,
  • The role of mental health, sleep, and physical activity in cognitive functioning,
  • The effects of digital distractions (social media, streaming) and part-time employment on academic focus,
  • The relevance of socio-demographic factors (e.g., gender, parental education) and lifestyle choices (e.g., diet, internet quality).

Our dataset captures diverse student profiles (ages 17–24) with metrics such as daily study hours, sleep duration, self-rated mental health, exercise frequency, and extracurricular engagement. Preprocessing ensured robustness, including handling missing data via exclusion and encoding categorical variables for analysis.

Preliminary Insights

Initial findings reveal that while some factors (e.g., gender, part-time jobs) show negligible effects, others demonstrate significant associations:

  • Study hours exhibit the strongest positive correlation with exam scores, reinforcing the centrality of dedicated learning time.
  • Mental health and sleep duration emerge as meaningful contributors to performance.
  • Frequent exercise, particularly at high intensity, correlates with improved outcomes.
  • Conversely, excessive social media and streaming hours impose small but consistent penalties on achievement.

This report details our methodology, validates findings through statistical testing, and culminates in evidence-based recommendations for students and educators. Subsequent sections outline data characteristics (Section 2), preprocessing steps (Section 3), exploratory visualizations (Section 4), statistical results (Section 5), and strategic implications (Section 6). By synthesizing these insights, we aim to empower stakeholders with actionable strategies to enhance academic success through targeted lifestyle interventions.

Key Terms: Academic Performance, Lifestyle Habits, Statistical Analysis, Correlation, Predictive Factors, Student Well-being.

2. Dataset Description

This dataset comprises responses from 1,000 students aimed at exploring the relationship between various lifestyle, demographic, and psychological factors and academic performance, measured through exam_score. The dataset includes 15 variables (excluding the target variable), spanning categorical, binary, and continuous types. Below is a detailed description of each variable:

Variable Descriptions

  1. age (Continuous)
    • Description: Age of the student in years.
    • Stats: Mean = 20.50, Std = 2.31, Min = 17, Max = 24
  2. gender (Binary)
    • Description: Gender of the student (0 = Male, 1 = Female).
    • Stats: 958 non-missing values, roughly balanced (Mean = 0.50)
  3. study_hours_per_day (Continuous)
    • Description: Average number of hours spent studying each day.
    • Stats: Mean = 3.55 hrs, Std = 1.47, Min = 0, Max = 8.3
  4. social_media_hours (Continuous)
    • Description: Average number of hours spent on social media per day.
    • Stats: Mean = 2.51 hrs, Std = 1.17, Min = 0, Max = 7.2
  5. netflix_hours (Continuous)
    • Description: Average number of hours spent watching Netflix per day.
    • Stats: Mean = 1.82 hrs, Std = 1.08, Min = 0, Max = 5.4
  6. part_time_job (Binary)
    • Description: Whether the student has a part-time job (0 = No, 1 = Yes).
    • Stats: 21.5% have part-time jobs (Mean = 0.215)
  7. attendance_percentage (Continuous)
    • Description: Percentage of classes attended.
    • Stats: Mean = 84.13%, Std = 9.40, Min = 56%, Max = 100%
  8. sleep_hours (Continuous)
    • Description: Average number of sleep hours per night.
    • Stats: Mean = 6.47 hrs, Std = 1.23, Min = 3.2, Max = 10
  9. diet_quality (Ordinal Categorical: 0 = Poor, 1 = Average, 2 = Good)
    • Description: Self-rated quality of diet.
    • Stats: Mean = 1.19, indicating diet quality tends toward “Average”
  10. exercise_frequency (Ordinal Categorical: 0–6)
    • Description: Number of days per week the student exercises.
    • Stats: Mean = 3.04 days, Std = 2.03
  11. parental_education_level (Ordinal Categorical: 0 = None, 1 = High School, 2 = College)
    • Description: Highest education level attained by the student’s parents.
    • Stats: Available for 909 students. Mean = 0.75
  12. internet_quality (Ordinal Categorical: 0 = Poor, 1 = Moderate, 2 = Good)
    • Description: Student’s assessment of their home internet quality.
    • Stats: Mean = 1.29, skewed toward “Moderate to Good”
  13. mental_health_rating (Ordinal Categorical: 1–10)
    • Description: Self-rated mental health on a 1 (low) to 10 (high) scale.
    • Stats: Mean = 5.44, Std = 2.85, Min = 1, Max = 10
  14. extracurricular_participation (Binary)
    • Description: Whether the student participates in extracurricular activities (0 = No, 1 = Yes).
    • Stats: 31.8% participate (Mean = 0.318)
  15. exam_score (Continuous)
    • Description: Final exam score, used as the outcome variable.
    • Stats: Mean = 69.60, Std = 16.89, Min = 18.4, Max = 100

Missing Data Note

  • parental_education_level has 91 missing entries.

These missing values were handled appropriately during the analysis phase to maintain the integrity of statistical results.

3. Data Preparation and Cleaning

Before conducting the analysis, several data preprocessing steps were carried out to ensure data quality and consistency:

  • Handling Missing Values: A total of 91 records with missing values in the Parental Education Level field were removed to preserve the validity of group comparisons.
  • Categorical Encoding: Categorical variables such as Gender and Parental Education Level were converted into numerical formats using label encoding or dummy variables to facilitate statistical testing and correlation analysis.
  • Irrelevant Column Removal: Non-informative columns like Student ID were dropped to streamline the dataset and avoid potential biases in analysis.
  • Data Type Verification: All variables were reviewed to ensure appropriate data types (e.g., numeric vs. categorical), and any inconsistencies or formatting issues were corrected.

These steps helped create a clean, analysis-ready dataset suitable for hypothesis testing and statistical modeling.

4. Exploratory Data Analysis (EDA)

General Distributions

Histograms

png

Count plots

png

Relationship with Exam Score

Scatter plots:

png

Boxplots:

png

Correlation matrix heatmap:

png

5. Statistical Analysis

Correlation Analyses

To examine continuous predictors of exam performance, Pearson correlation coefficients were calculated between exam scores and several continuous variables. The strength and direction of each relationship, as well as statistical significance, are described below.

Mental Health Rating

A moderate positive correlation was observed between students’ mental health ratings and exam performance (r = 0.318, p < 0.001). This indicates that better self-reported mental health is associated with higher exam scores, suggesting that psychological well-being may play a supportive role in academic success.

Age

There was no significant relationship between age and exam scores (r = -0.013, p = 0.699), suggesting that age does not predict performance in this student sample.

Study Hours

The strongest observed correlation was between study hours and exam scores (r = 0.823, p < 0.001). This represents a very strong positive association, confirming that increased study time is strongly linked to better academic performance. This is consistent with well-established educational findings and highlights the importance of sustained study habits.

Social Media Hours

A small but significant negative correlation was found between social media usage and exam performance (r = -0.172, p < 0.001). This suggests that more time spent on social media is associated with lower exam scores, possibly due to time displacement or reduced concentration.

Netflix Hours

A similar pattern was observed for time spent watching Netflix, with a negative correlation to exam performance (r = -0.167, p < 0.001). Like social media, this suggests that excessive screen time for entertainment may be detrimental to academic outcomes.

Attendance

There was a small but statistically significant positive correlation between attendance and exam scores (r = 0.096, p = 0.004). This suggests that students who attend more classes tend to perform slightly better, though the effect size is modest.

Sleep Hours

Lastly, a small but significant positive relationship was found between sleep duration and exam scores (r = 0.122, p < 0.001). This implies that getting more sleep is associated with improved academic performance, reinforcing the importance of sleep hygiene in student success.

T-Tests

Part-Time Job Status

To determine whether students’ exam scores differ based on part-time job status, both an independent samples t-test (Welch’s correction applied due to violation of normality) and a non-parametric Mann-Whitney U-test were conducted. The t-test yielded a t-statistic of 0.515 with a p-value of 0.607, and the Mann-Whitney U-test yielded a U-statistic of 71,900.500 with a p-value of 0.534. Both p-values are well above the conventional significance level of 0.05, indicating no statistically significant difference in exam scores between students with and without a part-time job. These results suggest that part-time employment is not associated with a meaningful difference in academic performance in this sample.

Gender

An independent samples t-test and Mann-Whitney U-test were also conducted to evaluate whether exam scores differ by gender. The t-test produced a t-statistic of -0.874 with a p-value of 0.382, while the Mann-Whitney U-test resulted in a U-statistic of 91,796.000 and a p-value of 0.352. As with part-time job status, these findings are not statistically significant. Therefore, no evidence was found to suggest a difference in exam performance based on gender in this dataset.

Extracurricular Participation

Lastly, the relationship between extracurricular involvement and exam scores was examined. The t-test produced a t-statistic of -0.103 with a p-value of 0.918, and the Mann-Whitney U-test yielded a U-statistic of 88,898.500 with a p-value of 0.851. Both tests strongly indicate a lack of significant association between participation in extracurricular activities and exam performance.

Group Comparisons Using ANOVA and Kruskal-Wallis Tests

Diet Quality

To assess whether exam performance varies based on students’ diet quality, both an ANOVA and Kruskal-Wallis test were conducted. The ANOVA returned F = 0.792 with p = 0.453, and the Kruskal-Wallis test returned H = 1.623 with p = 0.444. Both results indicate no statistically significant differences in exam scores across diet quality categories. Thus, diet quality does not appear to be associated with exam performance in this sample.

Exercise Frequency

In contrast, results for exercise frequency showed a significant association with exam performance. The ANOVA yielded F = 5.546 with p < 0.001, and the Kruskal-Wallis test returned H = 28.839 with p < 0.001, both indicating statistically significant differences in exam scores across exercise frequency levels.

To further explore which groups differed, a Dunn’s post hoc test with Bonferroni correction was performed. Key findings include:

Students who exercised daily (6 days/week) scored significantly higher than those who:

  • Did not exercise at all (0 days/week) (p = 0.004)
  • Exercised once per week (p < 0.001)
  • Exercised 3 days/week (p = 0.049)

The comparison between once per week and 5 days/week was also significant (p = 0.018).

These results suggest that more frequent exercise is positively associated with exam performance, with the most notable gains occurring at the highest frequency of physical activity.

Parental Education Level

Group comparisons based on parental education level revealed no significant differences in exam scores. The ANOVA yielded F = 0.942, p = 0.390, and the Kruskal-Wallis test returned H = 2.991, p = 0.224. These results imply that the educational attainment of students’ parents did not have a measurable impact on academic performance in this context.

Internet Quality

Finally, the impact of internet quality on exam performance was examined. Although some evidence of non-normality was observed in earlier diagnostics, neither the ANOVA (F = 1.518, p = 0.220) nor the Kruskal-Wallis test (H = 3.060, p = 0.217) found significant differences across internet quality levels. This suggests that, within this sample, perceived internet quality did not significantly influence exam outcomes.

6. Final Summary and Recommendations

This study explored a range of personal, behavioral, and contextual factors to understand what influences students’ academic performance, as measured by exam scores. By applying a combination of hypothesis testing, group comparisons, and correlation analyses, several key insights emerged.

Key Findings

  • Study habits matter most: The strongest predictor of exam performance was the number of hours spent studying (r = 0.823, p < 0.001). This robust association highlights the critical role of consistent and focused study in achieving academic success.
  • Mental health and sleep contribute meaningfully: Better self-rated mental health (r = 0.318) and adequate sleep duration (r = 0.122) were both positively associated with higher exam scores, suggesting that well-being and rest should not be overlooked as academic assets.
  • Exercise may help, especially at higher frequencies: Students who exercised more frequently (especially 6 times per week) outperformed those with little to no exercise. While overall effect sizes were modest, this supports the idea that regular physical activity may enhance cognitive performance or mood in a way that benefits learning.
  • Digital distractions hinder performance: Time spent on social media (r = -0.172) and Netflix (r = -0.167) showed small but significant negative relationships with exam performance. This suggests that excessive screen time may interfere with study time or focus.
  • Other demographic and contextual factors had little effect: Variables such as part-time job status, gender, extracurricular participation, parental education, diet quality, internet quality, and age showed no significant or meaningful relationships with exam performance in this dataset.

Recommendations

Based on these findings, the following evidence-based recommendations are offered for students and educators:

For Students:

  1. Prioritize consistent study routines: Regular and focused study sessions are the most powerful predictor of academic performance. Aim for daily dedicated time, and minimize multitasking.
  2. Protect your mental health: Seek support when stressed, maintain a balanced routine, and use mental health resources offered by your institution.
  3. Establish healthy sleep habits: Aim for at least 7–9 hours of quality sleep each night. Avoid late-night screen time and build a calming bedtime routine.
  4. Incorporate physical activity: Engage in moderate exercise most days of the week. Even a short daily walk or workout can contribute positively to academic outcomes.
  5. Limit digital distractions: Set boundaries around social media and streaming time—consider app timers or study apps like Forest or Pomodoro tools to stay on task.

For Educators and Institutions:

  1. Promote study skills and time management workshops: Equip students with strategies to study more effectively and manage distractions.
  2. Support student well-being: Offer accessible mental health resources and encourage practices that promote resilience and stress management.
  3. Encourage a balanced student lifestyle: Integrate wellness initiatives, exercise programs, and sleep hygiene education into student success programming.
  4. Monitor digital engagement: Consider educational campaigns that raise awareness about the impact of excessive screen time on academic focus and performance.