[R] Restricted fitting of two-component mixture distribution in R possible?

Johannes Moser jzmoser at gmail.com
Tue Jul 15 12:28:13 CEST 2014


Dear list,

In fitting a two-component Student's t mixture distribution to some data 
(standardized GARCH residuals) one of the components has an estimated 
degree of freedom of 0.6. This means that even the first moment of the 
mixture distribution would not exist.

The gamlss.mx package in R is used for estimation. gamlss.control, 
glim.control and MX.control seem not to support this kind of option -- 
or I couldn't find out how...

Is there a way to restrict the degree of freedom parameter estimates to 
be larger than, say, 3?

Many thanks in advance,
Johannes



More information about the R-help mailing list