Re: OMEGA HAS A NONZERO BLOCK

From:Leonid Gibiansky Subject:Re: [NMusers] OMEGA HAS A NONZERO BLOCK Date:Tue, 08 Oct 2002 08:47:02 -0400 Nick, I can suggest a case where the correlation between CL and V is 1. Suppose that you have high variability in bioavailability. Then all the parameters are scaled by the bioavailability. You have a choice of putting ETA to bioavailability or using correlated CV, V with correlation equal to 1. I've seen the data where these problems were equivalent (in terms of diagnostic plots, OF, etc.). In fact, I recovered the solution looking on the NONMEM results that gave me consistent CV-V correlation that was close to 1. I ended up using F1=1*EXP(ETA), but this does not disqualify the example. You may argue that more data would be needed to create a valid model, but there were no more data available. I hope you would not argue that correlation 0.5 would not be inappropriate in this case? Leonid