From: Jenny <zheng.22@osu.edu>

Subject: Error model

Date: Thu, 21 Jan 1999 11:53:50 -0800

Dear NONMEM users,

I am trying to build a Statistical Model. When I used additive model, the residual plot wres vs. pred is V-shaped, i.e., wider wres at high pred. When I used CCV error model, the plot is a reversed V-shape, i.e., wider wres at low pred. Therefore, I tried the power function model. The objective function did decreased, but the residual plot show a linear relationship, where the wres is more negative at high pred, wres is more positive at low pred. So far, I haven't got a good error model. If anyone has any suggestions, you are highly appreciated.

From: Jenny <zheng.22@osu.edu>

Subject: Re: Error model

Date: Thu, 21 Jan 1999 13:39:23 -0800

As I mentioned before, I have tried power function model. Actually, before power function model, I also tried combination of additive and CCV model. The combination model did not decrease as much as power model, the residual plot showed similar patten as CCV model alone. Does anybody think there are may be other problems associate with it? Logically, Any of this two models should do the work.