time_series

Time Series Scripts

Student worked figures

Run the master visualization script from the repository root:

python scripts/time_series/time_series_student_visualizations.py

It prints the numerical checkpoints used in the notes and writes the new figures to assets/time_series/student/. Existing figures in assets/time_series/ are retained and reused by the notes.

Topic	Existing example	Worked-figure script
Foundations	white_noise.py, random_walk.py	time_series_student_visualizations.py
Dependence and ARMA	autocorrelation_function.py, autoregressive_model.py, moving_average.py	time_series_student_visualizations.py
ARIMA and seasonality	arima_inflation.py, sarima_mock.py	time_series_student_visualizations.py
Forecasting	forecast_backtesting.py, exponential_smoothing_temp.py	time_series_student_visualizations.py
Dynamic and multivariate models	dynamic_regression.py, var_and_cointegration.py	time_series_student_visualizations.py
State space and frequency	kalman_filter.py, frequency_domain.py	time_series_student_visualizations.py
Financial volatility	sma_ema_stocks.py	time_series_student_visualizations.py

The scripts that fit statistical models need the dependencies listed in the repository requirements. The worked-figure script intentionally keeps its calculations explicit and only requires NumPy and Matplotlib.

Topic visualization programs

The expanded student chapters use one script per topic. Each program computes the numerical examples printed in its terminal output and writes eight figures to a topic directory.

Visualization script	Main lesson	Figure output
foundations_visualizations.py	series, recursions, components, moments, sampling, and stationarity	assets/time_series/foundations/
dependence_visualizations.py	covariance, ACF/PACF, AR/MA persistence, backshift, invertibility, and randomness	assets/time_series/dependence/
arima_seasonality_visualizations.py	differencing, decomposition, ARIMA forecasts, seasonal naive, and order selection	assets/time_series/arima_seasonality/
diagnostics_visualizations.py	detrending, residual checks, variance, breaks, and temporal splits	assets/time_series/diagnostics/
forecasting_evaluation_visualizations.py	baselines, rolling origins, intervals, metrics, leakage, and horizon error	assets/time_series/forecasting/
dynamic_multivariate_visualizations.py	dynamic regression, VAR feedback, Granger predictability, VECM, and impulses	assets/time_series/dynamic_multivariate/
state_space_frequency_visualizations.py	Kalman filtering, missing data, spectra, leakage, aliasing, and coherence	assets/time_series/state_space_frequency/
financial_time_series_visualizations.py	returns, ARCH/GARCH, leverage, persistence, and heavy tails	assets/time_series/financial/

Run any script from the repository root, for example:

python scripts/time_series/forecasting_evaluation_visualizations.py

The scripts use deterministic random seeds. They are teaching simulations rather than fitted analyses of a real data source, so changing the seed or sample length is a useful exercise.

Runnable examples for notes/time_series.

The directory progresses from white noise, random walks, stationarity, ACF and AR models through smoothing/ARIMA, then to forecast backtesting, dynamic regression, VAR/cointegration, state-space filtering, and frequency-domain analysis. For evaluation, prefer the temporal backtesting examples over shuffled cross-validation.