Applied Time Series Analysis and Forecasting

This course offers a comprehensive, practice-oriented foundation in forecasting and time series analysis for data science. It combines statistical principles with modern applied techniques, balancing conceptual understanding with hands-on work in R. Students will learn how to analyze temporal data, design forecasting workflows, and evaluate predictive performance. By the end of the semester, students will be able to build and adapt forecasting models across a wide range of settings, evaluate them with principled metrics, and communicate their results effectively. The emphasis is on balancing accuracy with interpretability, and on producing forecasts that are both rigorous and directly useful for decision-making in real-world context.

Student Learning Outcomes

1. Define and distinguish fundamental concepts in time-series and forecasting including data structures, time series components (trend, seasonality, cycles),  and major model families, and recognize core challenges such as nonstationarity, time-varying dynamics, and high-dimensional predictors.
2. Explain the theoretical foundations and key assumptions of major time series models (e.g., ETS, ARIMA, VAR/VECM, ML approaches) and justify how different model classes are selected or adapted to address specific data and decision contexts.
3. Wrangle and transform time series datasets; conduct rigorous exploratory data analysis (EDA); and implement appropriate forecasting models while using diagnostic checks and evaluation metrics to assess specification quality and predictive performance.
4. Critically evaluate and compare alternative model specifications by examining their assumptions, accuracy, robustness, interpretability, and operational constraints, and defend methodological choices using statistical evidence and domain knowledge.

Previously listed as DATASCI W271. Until 2025, the course was titled “Statistical Methods for Discrete Response, Time Series, and Panel Data”.