[R-SIG-Finance] FW: Re: racd package

[Trung.BA]

Hi Alexios and team, 

I am using racd in modelling time varying higher moments for a daily return
of SP500 since 2000. 

I tried to used msoptim solver with n.sim = 10000. I am now considering
skewed t ("sstd"), NIG and SGED distribution.

I got this warning message :" In .acdmakefitmodel(acdmodel = fname, f = fun,
T = T, m = m, timer = timer,  :

  NaNs produced"

Then, when I plot the skewness and kurtosis parameter from estimated models,
they varies in the similar manner but in a large different value. For
example, kurtosis in sstd distribution is between 1 - 100, kurtosis of NIG
is even between 1-300. 

Is there anything wrong with my estimation that relate to surprising
differences in kurtosis and skewness estimated value? 

Thanks in advance 

Trung


	[[alternative HTML version deleted]]