tag:blogger.com,1999:blog-7905515.post356815519209615200..comments2023-01-28T05:19:44.944-06:00Comments on Falkenblog: One-Month Trading StrategiesEric Falkensteinhttp://www.blogger.com/profile/07243687157322033496noreply@blogger.comBlogger1125tag:blogger.com,1999:blog-7905515.post-80174754488375520362022-06-07T13:11:16.944-05:002022-06-07T13:11:16.944-05:00"In contrast, multivariate regression weights..."In contrast, multivariate regression weights (coefficients) will be influenced by the covariances among the factors as well as the correlation with the explanatory variable."<br /><br />You don't need regression, per se. It obviously depends on assumptions you are making, but in the simplest case you just need expected returns of each factor. From this, you can produce expected returns of each stock. if you assume that the expected returns of each factor are equal, then the resulting expected returns of each stock are just a linear transformation of the sum of the individual factor scores.<br /><br />Anyway, if regressions are a mess, use regularization or Bayesian techniques to impose a prior. Johnhttps://www.blogger.com/profile/01457388998903348000noreply@blogger.com