Enji_ProblemSet1_STATS201

Problem Set 2 – Revised Proposal

Acknowledgment

I am grateful to Professor Luyao Zhang and my peer reviewer, Undran Enkhbaatar, for thoughtful and constructive feedback on Problem Set 1. Their comments helped me sharpen the research question, justify methodological choices with credible sources, and improve the coherence of the narrative.

---

Point-by-Point Response to Feedback

• Professor Zhang: Requested clearer definitions, novelty statement, reproducibility, and reflexivity.
  ✅ Added method definitions, cited best practices, clarified contribution, documented imputation/weighting, and included uncertainty bands.

• Peer Reviewer (Undran): Asked why equal vs PCA weighting, how imputation bias is handled, and requested figures.
  ✅ Explained equal vs PCA rationale, tested both imputation methods, and committed to visualizations (time-series plots, heatmap, forecasts).

---

Background and Motivation

• Composite indicators are sensitive to scaling, imputation, weighting → need transparency (OECD 2008).
• ESG research shows divergence across methodologies (Berg, Kölbel, and Rigobon 2022).
• Mongolia: highly energy-intensive, lagging renewable uptake, but 2030 policy targets for efficiency & diversification.
• China: benchmark case of successful energy intensity reduction and renewable scale-up (IEA 2020).

---

Research Question

How have Mongolia’s renewable energy share and energy intensity evolved since 1990, how sensitive are conclusions to alternative imputation and weighting choices in a simple mini-index, and what lessons from China’s experience could inform Mongolia’s path to meeting its 2030 energy policy goals?

---

Literature Review

• Composite indicators: need transparency, sensitivity checks (OECD 2008).
• ESG ratings divergence motivates robustness (Berg et al. 2022).
• EPI audits show aggregation/weighting can reshape narratives (Wendling et al. 2020).
• IEA: China’s efficiency programs → benchmark for Mongolia.

---

Indicators

• Renewable Energy Share (EG.FEC.RNEW.ZS) – measures diversification from coal.
• Energy Intensity of GDP (EG.EGY.PRIM.PP.KD) – SDG 7.3 efficiency indicator.
• Both are World Bank indicators → reproducible and comparable.

---

Methodology

1. Data & Preprocessing

   • World Bank (1990–2021).
   • Imputation: forward-fill for short gaps, mean for long gaps.
   • Standardized to z-scores.

2. Mini-Index Construction

   • Equal weights (transparent baseline).
   • PCA weights (variance-driven alternative).
   • Compare robustness across imputation + weighting choices.

3. Forecasting (to 2030)
   Models applied separately to each indicator:

   • ARIMA (Box–Jenkins).
   • Random Forest (nonlinear).
   • AutoGluon-Tabular (AutoML ensemble).
   • Metrics: RMSE & MAE.
   • Forecasts: 2026–2030 with prediction intervals.

---

Anticipated Results & Visualizations

• Energy intensity: gradual decline but still high.
• Renewable share: modest growth from low base.
• Index differences: slope/level shifts highlight sensitivity.

Figures planned:

• Time-series plots (1990–latest).
• Sensitivity heatmap (weighting vs imputation).
• Forecast plots (2026–2030 with uncertainty bands).

---

Ethical & Practical Considerations

• Descriptive (not causal).
• Transparent about assumptions & uncertainty.
• FAIR: findable & accessible, but limited interoperability.
• CARE: avoid deficit framing, acknowledge governance biases.

---

References

• Berg, Florian, Julian F. Kölbel, and Roberto Rigobon. 2022. Aggregate Confusion: The Divergence of ESG Ratings. Review of Finance 26 (6): 1315–44.
• Box, George E. P., Gwilym M. Jenkins, and Gregory C. Reinsel. 2015. Time Series Analysis: Forecasting and Control. 5th ed.
• Breiman, Leo. 2001. “Random Forests.” Machine Learning 45 (1): 5–32.
• Erickson, Nick, et al. 2020. AutoGluon-Tabular: Robust and Accurate AutoML for Structured Data. arXiv:2003.06505.
• IEA. 2020. Energy Efficiency 2020. Paris: IEA.
• Jolliffe, Ian T., and Jorge Cadima. 2016. “Principal Component Analysis: A Review and Recent Developments.” Philosophical Transactions of the Royal Society A 374.
• OECD. 2008. Handbook on Constructing Composite Indicators. Paris: OECD.
• Wendling, Z.A., et al. 2020. 2020 Environmental Performance Index. Yale Center for Environmental Law & Policy.

---

flowchart TD
  A[Background & Motivation] --> B[Research Question]
  B --> C[Indicators]
  C -->|Renewable Share + Energy Intensity| D[Methodology]
  D --> D1[Data & Preprocessing]
  D1 --> D2[Mini-Index Construction]
  D2 --> D3[Forecasting to 2030]
  D3 --> E[Anticipated Results & Visualizations]
  E --> F[Ethical & Practical Considerations]
  F --> G[Policy Implications: Lessons from China]

  %% cross-links / annotations
  A -.->|Comparative context| B
  D2 -.->|Equal vs PCA Weights| E
  D3 -.->|ARIMA / ETS / DirectTabular| E

Loading

System Configuration

• Local: Python 3.12, Jupyter Notebook, pandas, scikit-learn, statsmodels, matplotlib, seaborn
• Cloud: Google Colab (GPU optional)
• Version Control: Git + GitHub
• Reproducibility: Deterministic seeds, requirements.txt to be added