Time series analysis and forecasting
Hands-on implementation of classical time series models as part of advanced MSc coursework, focused on understanding how these models actually behave rather than just knowing the theory.
Understand how classical time series models behave in practice: when their assumptions hold, when they break, and how to diagnose failures from the data rather than from textbook descriptions.
Implemented AR, MA, ARIMA, and ARIMAX models from scratch, including stepwise model selection procedures. Ran full diagnostic suites: residual analysis, autocorrelation checks, and heteroskedasticity tests (ARCH effects). Explored model selection criteria and the trade-offs between model complexity and generalization.
Compared model behavior across different datasets and applied hold-out evaluation to distinguish in-sample fit from out-of-sample performance. Checked that residuals satisfied model assumptions before accepting any fit as valid.
A much more grounded understanding of time series modeling and its limitations on real data, particularly around non-stationarity, seasonal effects, and when classical models are and are not appropriate.