Practitioner course in quantitative foundations
Time Series Analysis for Finance
The time series methods behind trading signals and risk forecasts: stationarity, autocorrelation, ARMA models, unit roots, volatility clustering and honest forecast evaluation.
Financial data arrives as time series, and the order matters. Yesterday’s volatility predicts today’s; a spread that drifts too far tends to come back; a model fitted to calm years fails in a crisis. Time series analysis is the set of methods that take the ordering seriously.
What is time series analysis used for in finance?
Forecasting volatility for risk and option pricing, modelling spreads for statistical arbitrage, testing whether a signal leads returns, and measuring how relationships between markets change. Each application appears in the course on real data.
Who is this course for?
Practitioners building trading signals or risk forecasts, and students preparing for statistical arbitrage or machine learning courses. It counts toward the Certificate in Statistical Arbitrage and the Certificate in Machine Learning for Trading.
What you will be able to do
- Test a price or return series for stationarity and transform it appropriately.
- Identify, fit and diagnose ARMA models for returns and spreads.
- Recognise volatility clustering and fit a first GARCH model.
- Test for unit roots and explain why they matter for spread trading.
- Evaluate forecasts out of sample against a naive benchmark.
Syllabus, module by module
Time series data
Prices, returns, sampling frequency, missing data and alignment across markets.
Stationarity and autocorrelation
Weak stationarity, the autocorrelation and partial autocorrelation functions, and their sampling error.
ARMA models
Autoregressive and moving average models, identification, estimation and residual checks.
Unit roots
Random walks, the Dickey-Fuller and KPSS tests, and differencing.
Volatility clustering
Why squared returns are autocorrelated even when returns are not; a first ARCH and GARCH fit.
Multivariate series
Vector autoregressions, Granger causality and lead-lag relationships between markets.
Structural breaks
Detecting changes in mean and variance, and what they do to fitted models.
Forecast evaluation
Rolling-origin evaluation, loss functions and the Diebold-Mariano test.
Project lab
Supervised work on the assessed forecasting project.
How the certificate is earned
An assessed project: you forecast a financial series of your choice, compare at least three models against a naive benchmark out of sample, and submit the code with a short report. On a pass you receive the Course certificate in Time Series Analysis for Finance, issued by the Institute with a verification code. Read how certification works.
Before you start
Statistics to regression level and working Python with pandas.
This course counts toward
- Certificate in Statistical ArbitrageFive courses and a capstone on market-neutral trading: time series foundations, pairs and baskets, cointegration and the validation that keeps spread strategies honest.
- Certificate in Machine Learning for TradingFive courses and a capstone on applying machine learning to markets without fooling yourself: labelling, features, purged validation, overfitting tests and reinforcement learning.
Questions applicants ask
Can time series models predict stock prices?
Daily stock returns show very little linear predictability, and the course demonstrates this. Time series models are more useful for volatility, spreads and risk, which is where most of the course’s applications lie.
Does the course use R or Python?
Python, with statsmodels and arch.