I'm trying to estimate a bivariate GARK BEKK with the mgarchBEKK R pack, but that does not work. The problem is that "H is singular
The error is:
Error in buff.par.transposed[[tmp.count + 1]]% *% as.matrix (eps[count -: non-compliant arguments In addition: Warning message: In BEKK (diferenca.taxas1mo, order = c (1, 1), method = "BFGS"): negative inverted hessie matrix element
My code is:
bekk.taxas1mo <-BEKK (diferenca.taxas1mo, order = c (1,1), method = "BFGS")
One thing that made me think is that the series I'm working on is about 1/4 zeros. Anyone know if it's a problem?
Is mgarchBEKK a good package to use?
Thank you in advance.