Applied Time Series, Part 1: Statistics and Econometrics of Data with Memory

Learn to analyze and forecast data that remembers its past. This introductory statistics and econometrics course builds the complete applied time series workflow in Python: the time series mindset (why temporal dependence breaks cross-sectional intuition, what memory and autocorrelation mean, and how look-ahead bias silently corrupts naive analysis), practical data wrangling and visualization with pandas, decomposition into trend, seasonality, and remainder, the classical exponential smoothing family, stationarity and its core mechanics (ACF/PACF, differencing, unit root tests, spurious regression), the stochastic building blocks — white noise, random walks, AR, MA, and ARMA processes — assembled into the full Box-Jenkins ARIMA workflow, seasonal SARIMA models and regression with ARIMA errors, and finally honest evaluation: time-aware train/test splits, rolling-origin cross-validation, MASE and friends, and the discipline of always beating a naive benchmark before claiming success. Students work throughout on four datasets — the classic airline passengers series, US macroeconomic series from FRED, daily electricity demand, and a synthetic suite of simulated processes with known parameters so every estimator can be scored against ground truth — and finish with an end-to-end forecasting capstone on held-out data. The organizing theme is that a time series is one realization of a process with memory: every method in the course is a way of measuring that memory, removing the parts that are predictable structure, and being honest about what remains.

  • Level: beginner
  • Duration: 10 weeks
  • 11 sections
  • Price: $50

Topics: time series, autocorrelation, ACF, PACF, lag plot, stationarity, unit root, ADF test, KPSS test, differencing, white noise, random walk, spurious regression, decomposition, STL, seasonal adjustment, moving average smoothing, exponential smoothing, Holt-Winters, AR process, MA process, ARMA, ARIMA, SARIMA, SARIMAX, Box-Jenkins methodology, Ljung-Box test, AIC, BIC, forecast horizon, prediction intervals, naive forecast, seasonal naive, MASE, rolling-origin cross-validation, backtesting, look-ahead bias, statsmodels, pandas

What's inside

  1. Applied Time Series, Part 1: Statistics and Econometrics of Data with Memory
  2. Chapter 1: Thinking in Time: Data with Memory
  3. Chapter 2: The Time Series Toolbox: Wrangling and Visualization in Python
  4. Chapter 3: Decomposition: Trend, Seasonality, and the Remainder
  5. Chapter 4: Classical Forecasting: The Exponential Smoothing Family
  6. Chapter 5: Stationarity and Core Mechanics: ACF, Unit Roots, and Spurious Regression
  7. Chapter 6: Building Blocks I: White Noise, Random Walks, and Autoregressive Models
  8. Chapter 7: Building Blocks II: Moving Average Models, ARMA, and Model Signatures
  9. Chapter 8: ARIMA: The Box-Jenkins Workflow
  10. Chapter 9: SARIMA and Regression with ARIMA Errors
  11. Chapter 10: Evaluation, Benchmarks, and the Forecasting Workflow — Capstone

Read a free sample chapter