I'm trying to estimate/fit a multivariate t distribution, setting the skew/shape parameter vector Alpha[8,1] to 1, but to no avail. The 'freq' option seems only to be for and n x 1 vector of weights, while I need one for each of 8 margins in the distribution. None of the options from the nlminb optimiser into which the problem is passed seem to work, either. Has anybody had experience restricting otherwise random parameters to constants? Casey Quinn.
Reasonably Related Threads
Wisdom of the Ancients
