Maybe I’m dense, but I still don’t understand how PCA analysis and regression can be used to actually choose stocks to buy and sell. Perhaps you (duckruck) could explain this? I have tried doing multilinear regression on factors and groups of factors and have failed to come up with anything at all, likely because I’m ignoring some essential step.
Here is what I understand so far. You can combine factors using PCA to come up with new factors. You can perform multilinear regression on, say, the returns of a long-short portfolio based on each of those factors against the returns of a benchmark. You would then come up with a formula consisting of, say, 6 betas and an alpha on the six PCA factors. But what do you do then? How do you actually use that to choose stocks? Whenever I’ve tried anything like this, the actual betas I’ve come up with are often strongly negative and have almost no statistical significance (extremely high p-values), and if you’re trying to choose stocks that are likely to outperform or underperform, I don’t see why you’d want to regress to a benchmark rather than simply favoring the long-short portfolios with the highest return. I’m 100% sure I’m doing something wrong, so I’d love to be corrected.