Statistical Analysis of Critical Minerals Demand in Clean Energy Industries
📅 Timeframe: September 2024 - December 2024
📍 Author: Jess Valiarovski
📚 Course: Applied Biostatistics BIOL 3272
🎓 Acknowledgements: Thank you, Yaniv Brandvain and Abigail Meyer!
The transition from fossil fuels to low or zero-carbon energy is a major challenge for clean energy industries.
This project examines how mineral demand varies across industries using linear regression, bootstrapping, and pairwise statistical tests to provide robust insights into demand patterns using R-programming.
- Test Hypothesis: Does mineral demand significantly differ across industries?
- Compare Statistical Methods: Linear modeling vs. resampling techniques.
- Apply Statistical Modeling: ANOVA, bootstrapping, and multiple comparisons to detect industry-specific trends.
- Source: International Energy Agency (IEA) Critical Minerals Survey 2023
- Data Type: Mineral demand in kilotons (kT) per industry
- Data Cleaning & Transformation:
dplyr,tidyverse,janitor - Statistical Modeling:
lm()for linear modeling and bootstrapping - Multiple Comparisons: Holm’s correction
- Effect Size Analysis: Cohen’s d
- Violin plots for demand distributions
- Bootstrapped confidence intervals to highlight sampling variability
- Pairwise comparisons with Holm correction to identify significant differences
📌 Industry sector influences mineral demand significantly (p = 0.0395 < 0.05, F = 2.9391).
📌 Electric Vehicles (EVs) have the highest demand, while Hydrogen Technologies have the lowest.
📌 Bootstrapped confidence intervals show that all industries significantly differ.
📌 Without bootstrapping, industry demand appears similar, likely due to low sample sizes.
1️⃣ Log-Transformed Demand Visualization
To reduce residual variance & improve linear model assumptions.
2️⃣ Industry Demand Variability & Confidence Intervals
ANOVA post-hoc comparisons using Holm’s correction to identify significant differences in demand.
🛠️ Technical Skills Demonstrated
✅ Data Wrangling: dplyr, tidyverse, janitor
✅ Statistical Modeling: lm(), anova(), emmeans()
✅ Resampling Techniques: boot, infer
✅ Data Visualization: ggplot2, violin plots, bootstrapped CI
✅ Effect Size & Significance Testing: Cohen’s d, Holm’s correction
✅ Hypothesis Testing: Type II ANOVA, multiple comparisons
📌 GitHub Pages Link:
📎 View the Full Report Here
🔹 Identify which minerals each industry uses the most (e.g., Copper, Selenium) and compare the data to global resource scarcity and the mining benefits and risks/costs.
🔹 Try out Bayesian methods to quantify uncertainty in demand estimates.