Different versions of mode_compute
Posted:
Tue Mar 25, 2014 4:20 pmby klg123
A general question -- why should some versions of mode_compute work while others result in an error about not having a positive definite matrix? For my code, for example, mode_compute=6 works while 4, 8, 9, and 10 do not. Is this an issue?
Re: Different versions of mode_compute
Posted:
Tue Mar 25, 2014 6:32 pmby klg123
In addition, is there any way to specify the starting jscale value for mode_compute=6?
Re: Different versions of mode_compute
Posted:
Wed Mar 26, 2014 8:54 amby jpfeifer
It generally means you have not found a true mode (characterized by a semi-definite Hessian). This is an issue if you do not run your chain until convergence.