Foundation course in quantitative foundations
Statistics for Finance and Trading
Estimation, testing and regression for financial data, with the traps that make trading research go wrong: dependence, fat tails and testing too many ideas.
Most trading ideas that fail in live trading did not fail because the market changed. They failed because the statistics that supported them were wrong: a standard error that ignored autocorrelation, a result that was the best of fifty tries, a regression driven by three days in a crisis. This course teaches statistics for finance: the methods that catch those mistakes.
What is quantitative analysis in finance?
It is the use of data and statistical models to support financial decisions: whether a strategy has an edge, how much a stock moves with the market, how volatile a portfolio is. The methods are standard statistics. The difficulty is financial data, which is noisy, dependent and short relative to the effects being measured.
Who is this course for?
Anyone who will build or review trading strategies, factor models or risk estimates. It is part of the Foundation Certificate in Computational Finance.
What you will be able to do
- Estimate means, volatilities and correlations with honest standard errors.
- Run and interpret hypothesis tests, and explain what a p-value does and does not say.
- Fit and diagnose linear regressions on financial data.
- Correct for multiple testing when many strategies or factors are tried.
- Use the bootstrap where textbook formulas assume too much.
Syllabus, module by module
Describing returns
Simple and log returns, summary statistics and the stylised facts of financial data.
Estimation
Estimators, bias, variance and standard errors; why expected returns are so hard to estimate.
Hypothesis testing
Tests, p-values, power and the difference between statistical and economic significance.
Linear regression
Least squares, inference on coefficients and the regression behind beta and factor models.
Regression diagnostics
Heteroskedasticity, autocorrelation, robust standard errors and outliers.
Resampling
The bootstrap and block bootstrap for dependent data.
Multiple testing
Why testing a hundred strategies guarantees a false winner, and the Bonferroni, Holm and false discovery rate corrections.
Putting it together
A full analysis of a candidate signal, from data cleaning to a defensible conclusion.
How the certificate is earned
Each module ends with a graded quiz, and the course ends with a graded problem set analysing a return data set in Python. On a pass you receive the Course certificate in Statistics for Finance and Trading, issued by the Institute with a verification code. Read how certification works.
Before you start
Probability at the level of Probability for Quantitative Finance, and enough Python to load a file and plot a series.
This course counts toward
Questions applicants ask
Which software is used?
Python with pandas, statsmodels and SciPy, in notebooks you can run in a browser.
Does the course cover time series models?
Only the basics needed for dependence. Time Series Analysis for Finance covers ARMA, GARCH and forecasting in depth.