On page 32 （PDF）
Two different results were obtained using the Global optimization.
Log likelihood does not differ much. BUT the estimates vary a lot, such as mu[p].
When I tried to use one of the answer from a particular run, I get the HFLOATING error, see picture.
So how reliable is this? Could there be a better way to optimize this ?
As an additional note, if I have Matlab R2014a, could I use Matlab to optimize the target function? DO I need to purchase a seperate addon?