search

jasonzy121 / Deep_Learning_Asset_Pricing

61 stars 47 forks source link

About SDF #2

Closed Ed-Chai closed 2 years ago

Ed-Chai commented 2 years ago

Excuse me, why M=1-sum(WR) in the paper while in the code implementation M=1+sum(WR) ？

jasonzy121 commented 2 years ago

Sorry for the confusion, this is a small inconsistency from an early draft of the paper, and you can think of w in code here is equivalent to -w in the paper

Ed-Chai commented 2 years ago

Thank you very much! Because I am new in this domain, some questions confusing me a lot! In step3, why could we generate RF=RF50? What's the meaning of multipling 50? And why could we approximate RF50 in step4 to get estimation of beta，and why could we use beta as weight to construct a long-short portfolio to calcultate sharpe ratio? I remember in factor model, weight=(beta.T.dot(beta))^-1.dot(beta.T). step5: why could we to get predictive R_t as beta_t.dot(R_t)/w_t.dot(w_t)*w_t?