Alternative factor models

QE008 · Quant Equity Investing · Module 8 of 8. Eleven interactive steps, each with its own guided explanation.

Four ways to build a factor model, compared on the same S&P 500 data: statistical factors from PCA and whether they hold up out of sample, one scoreboard for explaining risk, the q-factor model derived from how firms invest, spanning tests between models, machine learning read honestly (every model lost to a forecast of zero), and instrumented PCA, where betas follow the company.

Every step uses real market data. This is education about models and their assumptions, not financial advice.

Out-of-sample R² for five machine-learning models, all below zero, against the 0.4% published on a much larger panel.

Explore next

Stories, illustrations, and interactive mini lessons. Ask a question.