Re: Follow up: Model building advice

From: "Leonid Gibiansky" leonidg@metrumrg.com Subject: Re: [NMusers] Follow up: Model building advice Date: Thu, May 19, 2005 7:39 am You have too many random effects. Try to restrict it to 4-5, at least initially: CL,V,KE0,EMAX. If this would run, add new ones if needed. But the problem may be deeper: with new doses, you introduce jumps in the solution while gradient methods (that are in the foundation of nonmem search algorithms) and objective function approximations do not like jumps. You may need to look for a more smooth approximation of your jumps (something like a sharp sigmoid rather than a jump in the solution) if this is possible. Leonid _______________________________________________________