LOGMVSKEWT—function for multivariate skew T density
Posted: Thu Mar 22, 2012 10:49 am
This computes the (log) multivariate skew-t density from Bauwens & Laurent(2005), "A New Class of Multivariate Skew Densities, With Application to Generalized Autoregressive Conditional Heteroscedasticity Models," JBES, vol 23, pp 346-354. As this is a function (not a procedure), you will need to source the file in before you can use it.
Note that this is for a standardized density (mean 0, covariance matrix I). If your model generates mean zero data with covariance matrix <<sigma>>, you need to standardize the input vector by pre-multiplying by inv(%decomp(sigma)) and adjust the likelihood by subtracting .5 * log |sigma|.
%LOGMVSKEWT(z,xi,nu)
Parameters
[tr][td]Z[/td][td][/td][td]a K vector, which is assumed to represent a standardized (mean 0, covariance matrix I) process[/td][/tr]
[tr][td]xi[/td][td][/td][td]a K vector of (positive) skewness parameters (xi(i)=1.0 means no skewness, > 1 means skewed positive, < 1 means skewed negative)[/td][/tr]
[tr][td]nu[/td][td][/td][td]degrees of freedom for underlying Student t.[/td][/tr][/table]
Note that this is for a standardized density (mean 0, covariance matrix I). If your model generates mean zero data with covariance matrix <<sigma>>, you need to standardize the input vector by pre-multiplying by inv(%decomp(sigma)) and adjust the likelihood by subtracting .5 * log |sigma|.
%LOGMVSKEWT(z,xi,nu)
Parameters
[tr][td]xi[/td][td][/td][td]a K vector of (positive) skewness parameters (xi(i)=1.0 means no skewness, > 1 means skewed positive, < 1 means skewed negative)[/td][/tr]
[tr][td]nu[/td][td][/td][td]degrees of freedom for underlying Student t.[/td][/tr][/table]