Statistical analysis and predictive modeling of gym member fitness data.
Comprehensive exploratory and statistical analysis of 973 gym members covering:
- Physical attributes (age, weight, height, BMI, body fat)
- Exercise routines (type, duration, frequency, intensity)
- Performance metrics (calories burned, heart rate)
Exploratory Analysis
- Descriptive statistics and correlation matrices
- Principal Component Analysis (PCA) & Multiple Correspondence Analysis (MCA)
- Pattern identification via clustering (3 distinct member groups)
Statistical Testing
- Hypothesis tests (Shapiro-Wilk, Kolmogorov-Smirnov, Chi-squared)
- Normality and independence checks
- Gender & sport type dependency analysis
Clustering
- Hierarchical Agglomerative Clustering (HAC) with Ward linkage
- K-means clustering (K=3 optimal)
- Gaussian Mixture Models (Mclust)
- Cluster comparison (Adjusted Rand Index)
Predictive Modeling
- Multiple regression for calories burned (R² = 0.97)
- Two-way ANOVA with interactions
- ANCOVA with model selection (AIC/BIC)
- Logistic regression for experience level prediction
- Calories Model: Duration & heart rate are primary drivers (R² = 0.99 with interactions)
- Experience Level: Session duration is strongest predictor (p < 0.001)
- Clustering: Members separate into athletic high-performers + 2 morphology-based groups
- Gender Effect: No significant difference in training habits, only physical attributes
StatsCode.qmd- Quarto source code (R + Markdown)Rapport-Projet.pdf- Final report (25 pages max)- Data: 973 gym members × 13 variables
- R Packages: ggplot2, factoextra, mclust, stats, lme4
- Techniques: PCA, MCA, HAC, K-means, Mclust, linear/logistic regression, AIC/BIC selection
Boulaalam El-Mehdi, Ajerame Ilias, Grande Revuelta Ana
Course: Data Analysis & Statistical Modeling (ModIA UF)
Institution: INSA Toulouse
Date: January 2026
Note: Final report limited to 25 pages. Code available in .qmd file.