RE: FW: OMEGA HAS A NONZERO BLOCK

From:"Kowalski, Ken" Subject:RE: FW: [NMusers] OMEGA HAS A NONZERO BLOCK Date:Fri, 4 Oct 2002 10:57:46 -0400 Leonid, What happens when you use your proposed parameterization when Omega is ill-conditioned as in Steve's example? Since NONMEM is wanting to estimate the correlation between P1 and P4 (in your example CL and V2) to be 1 doesn't that mean that THETA(8)=THETA(9)=THETA(10)=0 (since ETA(1), ETA(2) and ETA(3) are assumed to be independent)? So, with your parameterization, I contend your model will reduce to 7 parameters in Omega as well. Ken