This code implements two factor extraction methods: the heteroskedastic factor analysis (HFA) of Jones (2001) and the asymptotic principal components procedure (CK) of Connor and Korajczyk (1986). The HFA method can improve the quality of the extracted factors by allowing for heteroskedastic residuals. Besides, it does not require any extra assumptions to those in CK. The user must specify a T*N matrix of returns and the number of factors to extract. The number of iterations in the optimization process is by default set to 5. The minimum allowable average ideosyncratic standard deviation is by default set equal to .001.
Working Paper (2018)
Müller Gernot and Hettig Thomas