Student Performance Regression Analysis
Investigated factors associated with academic performance using multiple linear regression, model diagnostics, and cross-validation in R.
Overview
A group project for Purdue University’s STAT 512 course investigating the relationship between academic performance and student characteristics.
Using R, we developed multiple linear regression models to analyze how study hours, previous academic scores, sleep duration, extracurricular activities, and practice questions relate to a student’s performance index.
Methodology
Data Preparation
- Standardized numerical variables for comparison.
- Encoded extracurricular participation as a binary variable.
- Prepared the dataset for regression analysis.
Regression Modeling
Constructed a multiple linear regression model using five explanatory variables:
- Hours studied
- Previous scores
- Extracurricular activities
- Sleep hours
- Sample question papers practiced
Model Diagnostics
Evaluated regression assumptions and potential limitations using:
- Residual analysis: Histograms and Q-Q plots to examine normality.
- Multicollinearity: Variance Inflation Factors (VIF).
- Influential observations: Cook’s distance and influence plots.
- Linearity: Added-variable plots.
- Variance: Breusch-Pagan testing and variance diagnostics.
Model Validation
Applied 10-fold cross-validation using R’s caret package and backward subset selection to evaluate the regression model.
Project Materials
The repository contains the group’s R scripts and final statistical report.