Suppose you have a list of stock characteristics—size, book-to-market, momentum—and you want to know which ones genuinely predict future returns. You cannot just look at average returns of portfolios sorted on each characteristic, because characteristics overlap and their effects get tangled. Regression lets you ask: “Holding everything else fixed, does this characteristic still matter?” But stock returns are noisy, and the data has a special structure—many stocks observed over many months. This chapter shows you how to run the standard regression that handles that structure, and how to get standard errors you can actually trust.