[R-SIG-Finance] cgarchfit (rmgarch package): cannot reconcile likelihood of a Copula-GARCH model
Ezequiel Antar
ezequ|e|@@nt@r @end|ng |rom gm@||@com
Fri Dec 11 11:14:31 CET 2020
I cannot reconcile the log-likelihood of a Copula-GARCH model.
Below is a simple example where the marginal volatilities are constant (for
which I use iGARCH with beta1 = 1) and have Normal distribution; and the
copula is constant and Gaussian. In short, the model is that of
returns having multivariate Normal distribution with constant parameters.
I compare the likelihood from cgarchfit (= 5890.564) with the likelihood
using the multinormal density, with both mean and covariance taken from
cgarchfit (= 5896.262). They should be the same.
Any help understanding why I cannot reconcile this will be much
appreciated!
-----------------------------------------------------------
library(rmgarch)
library(mvtnorm)
data(dji30retw)
z.t <- dji30retw[, 1:3]
uspec.each <- ugarchspec(mean.model = list(armaOrder = c(0,0)),
variance.model = list(model = "iGARCH", garchOrder
= c(1,1)),
distribution.model = "norm",
fixed.pars = list(beta1 = 1, alpha1 = 0, omega =
0))
mspec <- cgarchspec(uspec = multispec(replicate(ncol(z.t), uspec.each)),
distribution.model = list(copula = "mvnorm", method =
"Kendall", time.varying = FALSE, transformation = "parametric"))
cgarch <- cgarchfit(spec = mspec, data = z.t, solver.control=list(trace=1))
likelihood(cgarch) # 5890.564
mu <- coef(cgarch, type = 'garch')
cov <- rcov(cgarch)[,,1]
sum(dmvnorm(z.t, mean = mu, sigma = cov, log = TRUE)) # 5896.262
<https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail>
Virus-free.
www.avast.com
<https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail>
<#DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>
[[alternative HTML version deleted]]
More information about the R-SIG-Finance
mailing list